5.1 Sand events correlation
Based on the GPR image, the two strong reflectors identified were interpreted as the interfaces between the middle sand–peaty mud sequence with the upper tephra–mud sequence and the lower thick tephra (Fig. 3a). Further, the initial lateral correlation of the sand–peaty mud sequence was conducted based on sedimentary features, color, and stratigraphic position (Fig. 3c). The persistent presence of tephra layer four (T4) among several cores served as a marker horizon. Table 6 presents the lateral correlations among the core samples.
Table 6
Correlation of the sand layers among the core samples.
Sand layer | Correlation analysis |
T1 | It was defined as the Towada-a event (Table 3). It continues throughout all of the core samples (Fig. 3), just below the paddy field soil. It is also consistent that two more tephra events were found underlying it, interbedded with peaty mud, except in H13, H07, and H06, where there was only one, probably due to erosion. |
S1 | It only appears in the core sample H02 (the thickest sedimentary record). It is difficult to identify a lateral continuity, essentially due to its thickness (Table 1, Fig. 3). |
S2 | It was traced among H02, H01 and H07, mainly due to its relative stratigraphic position and lithological characteristic. |
S3 | It only appears in the core sample H02 (the thickest sedimentary record). It is difficult to identify a lateral continuity, essentially due to its thickness. |
S4 | It was traceable among H02, H01, H07, H08, and H03 (Fig. 3). Such correlation was made on the grain size, color and relative stratigraphic position. |
S5 | It was traced throughout all of the core samples, representing the sand with the furthest extent. It is the second thickest sand layer, after S8, in the sedimentary sequence. Its position just above the peaty-mud layer and T4 (marking horizon), upward fining, and bottom erosive contact allowed its lateral correlation. The dating result obtained from the peaty-mud layer underlying S5 between Ha8 and H04 and the supported the correlation (see the discussion section in the body text). |
T4 | It is a consistent ash layer, defined as a white clay with homogeneous thickness among the sand layers. Due to its consistency, lateral continuity, and lithological characteristics, it was used as a marker horizon to define the upper boundary of S6 and the lower boundary of S5, taking into account the peaty mud interbedding between the sand and tephra layers (Table 1, Fig. 3). |
S6 | It is located below the peaty mud underlying T4; it was found in the core samples H02 and H03. Such correlation was made based on the characteristic lenticular stratification, silt composition, and the thickness of the layers (Table 1). |
S7 | It was traceable among H02, H01, H07, H08, H03, H09, H10, and H04 (Fig. 3). Grain size, the bottom transitional or erosional contact, and the relative stratigraphic position led to its lateral correlation. |
S8 | It was traceable throughout all the core samples between H02 and H06 and represented, in general, the thickest sand layer. Its thickness, upward fining, and relative stratigraphic position allowed its lateral correlation. |
S9 | It was traceable among the core samples H02, H08, H03, H10, and H04. The two thin laminas interbedded with the thin peaty mud led to the correlation between H02 and H03. The stratigraphic position allowed the correlation among the other core samples. The dating result obtained from the peaty-mud layer underlying S9 and overlying S10 in Ha8 and H04 supported the correlation (see the discussion section in the body text). |
S10 | It was traceable among the core samples H02, H08, H03, H10, and H04. The stratigraphic position, overlying To-cu and the peaty mud, led to the lateral correlation. The dating result obtained from the peaty-mud layer underlying S9 and overlying S10 in Ha8 and H04 supported the correlation (see the discussion section in the body text). |
T6 | It was defined as the Towada-cu event (Table 3). It continues throughout all the core samples at the end of the sequence. In all of the cases, it was found as a thick lapilli layer. |
Table 6.
The correlation was confirmed and supported with CT, geochemical analysis, and radiocarbon dating, as will be described as follows.
5.1.1 Correlation between core samples H02 and Ha8
As the core sample Ha8 was the first core to be extracted, diatom analysis, tephra analysis, and radiocarbon dating for event recurrence was conducted on this core sample. The GPS coordinates indicated that the sampling site was the same for H02 and Ha8 (+/- 2 m).
The correlation with the adjacent H02 was determined based on the stratigraphic position of the sand layers, lithology, and sedimentary characteristics (Fig. 3, Table 1). Sand layers above S5 were correlated using the relative stratigraphic position because the lithological characteristics of the first four layers were identical. The mud layer above S5 in H02 can be correlated to the one overlying S5 in Ha8, with ages ranging from 2800 to 3000 cal BP and 3075 to 2973 cal BP, respectively.
In addition, T4 was below S5 in both cores, and it was used as a marker horizon (Fig. 3b). S6 was correlated based on the stratigraphic position below T4 and the sedimentary characteristics. The dating result between S8 and S7 in H02 did not coincide with the same stratigraphic position in Ha8. In contrast, such dating results coincided with the peaty-mud layer between S8 and S9 in Ha8. Nonetheless, the age difference is small, which could be related to the close time frame or reworking of sedimentary events during sand event deposition. However, it supports the correlation of S8 as it was deposited after 4892 cal BP, and the sedimentary characteristics of the layers coincide. In both cores, both S9 and S10 appear as thin beds underlying S8.
In general, the thickness of the sand layers changed considerably between the two core samples. S3 was not recognizable in Ha8, and S6 was significantly reduced in thickness to a medium lamina (Fig. 3b). Such fast lateral changes in thickness are common in tsunami deposits (Nakamura et al. 2012).
5.1.2 Lateral correlation of cores H02, H03, H04 and H06 using LDA and 14C
The ITRAX results and its analysis by LDA partially supported the initial correlation based on sedimentary features and stratigraphic relationships. LDA displayed a clear geochemical correlation of events S5 and S8 among core samples H02, H03, and H04 (Fig. 9). Nonetheless, events S5 and S8 in H06 cannot be explicitly linked to their corresponding levels in the other cores because of their closeness to other samples in the LDA classification. On the other hand, LDA evidenced a substantial geochemical similarity among the sands S6, S7, S9, and S10 in different cores, suggesting a similar lithological composition, and thus the source of the sediments. On the other hand, S4 in core H02 lacks a good correlation with the corresponding depth in H03 (Fig. 4), implying that the furthest landward sand could correspond to a different sedimentary event and source. Therefore, clustering must be made carefully to avoid the false correlation of sand layers. Nevertheless, LDA clustering is useful to find considerable differences among the samples, which suggests the need to confirm the reliability of the correlation, as seen in S4 (Fig. 9).
As for the correlation between Ha8 and H02, the 14C dating results supported the lateral correlation of the sand layers (Fig. 14). In addition, the mud layer overlying S10 in Ha8 can be correlated to the same stratigraphic layer in H04, with age ranges of 4892–5034 cal BP and 4854–4960 cal BP, respectively. Calibrated ages show that the peaty–mud layers underlying S5 in Ha8 and H04 are coetaneous. On the other hand, the dating result from the peat –mud underlying S7 in H04 does not correspond to the age obtained on the same stratigraphic level on Ha8. As stated previously, differences in some dating ages among core samples H02, Ha8, and H04 may be related to the reworking of the sampled material (Ishizawa et al. 2018), sediment erosion due to tsunami (Naruse et al. 2012), or sampling resolution (Fig. 14) (i.e., mud layers underlying S8 in H02 and Ha8 and mud layers underlying S7 in Ha8 and H04).
As the remaining cores (i.e., H13, H01, H07, H08, H09, H10, H11, H12, and H05) are intercalated with the core samples H02, H03, H04, and H06 (Fig. 3b), we utilized the results from radiocarbon dating, the position of T4, along with the sedimentary characteristics and stratigraphic position of the sand layers as critical elements for their lateral correlation.
5.2 Identification of tsunami origin
We consider marine diatom species, shell fragments, glauconite occurrence, landward thinning, and landward reduction of MS value as strong evidence of marine provenance of the sand layer (Table 2). Regarding marine provenance, a factual imprint is the presence of fragments or complete structures of marine organisms, which is, in our case, diatom species and shell fragments, respectively. Complementarily, glauconite also exhibits the marine origin of the sand layers by being an authigenic mineral of shallow oceanic platform environments (Huggett 2013).
Sand units S4, S5, S7, S8, S9, and S10 display landward reduction of its MS values among the core samples H02, H03, H04, and H06 (Fig. 5), indicating the direction of the density gradation during the wave progression. On the other hand, the maximum magnetic susceptibility for S6 was obtained from H02, which is located in the middle of the transect. (Figs. 5 and 7; Table 2). The MS value lateral comparison among the sand layers was used regardless of the sand layer thickness because it is related to the mineral content rather than its width.
Although not all sand layers have the same characteristics, sand units S4, S5, S6, S7, S8, S9, and S10 can be considered as clear marine-origin events, based on attributes such as landward thinning, erosive basal contacts, presence of diatoms of marine species, landward decrease in magnetic susceptibility value, upward finning, and glauconite occurrence (Table 2). Contrastingly, sand units S1, S2, and S3 lack any such evidence; they are not related to marine-origin events. Although the lack of marine evidence does not necessarily indicate a non-marine origin (Goff and Chagué-Goff 2012), its relationship to a tsunami event is weaker than that of the other sand units. Suppose a marine origin is assumed for sand units S1, S2, and S3. In this case, one possible explanation is that they were deposited close to the bore front, where the horizontal velocity was high, and the flow was considerably more turbulent. Therefore, the flow conditions only allow reduced sediment content, typically represented in mica flakes and clay minerals (Yoshii et al. 2017). Consequently, the remaining weak wave led to the deposition of very thin and extent-limited sand layers that were highly susceptible to leaching, weathering, and poor preservation of marine-related evidence.
Carbonate dissolution and XRD demonstrated that the positive calcium response to sand layers on ITRAX was related to the plagioclase content rather than the marine origin. The high presence of calcic plagioclase in the XRD results explains the calcium pattern in the ITRAX results (Appendix 1). Even though small quantities of carbonate were obtained in some sand layers (< 0.3 %) (Table 2), such values can be related to analytical error instead of real mineral content.
The presence of freshwater species related to the sand layers, mixed with marine species, can elucidate the reworking of marine provenance sediments with terrestrial material. Due to the minor thickness of the sand layers, it was not possible to determine whether the position of the marine species was related to a specific stratigraphic level. On the other hand, S8 exhibits that the marine diatoms are located in the base and top of the layer. This can be related to the increased mixing of marine and terrestrial materials during the progression and regression of the wave, in which the addition of in situ material is enhanced. Such behavior is not unusual because the background sedimentary environment is related to a marsh, where entrainment of freshwater species included in terrestrial sediments is expected. In addition, freshwater diatom species have been previously reported in the lithological units surrounding the study area (Kamada et al. 1991).
The grain size analysis was not entirely conclusive for inferring a marine origin. For instance, the grain size of the coarsest portion of the sand layers S4, S6, S7, S8, and S9 at H03 was finer than that at H02 (100 m seaward of H03). Horizontal changes in the grain size of the coarsest portion of the sand were not evident between H03 and H04, except for the sand layer S4. Thus, landward fining was not convincing among the core samples. This may be related to several factors, such as the particle size analysis performed with a small amount of sample, approximately 0.5 g each, the number of core samples is limited, and particularly because the changes in grain size are slight. F or instance, when comparing the maximum extent of the sand layer S8, approximately 200 m (Fig. 1 and Fig. 3b), to the number of core samples analyzed by grain size, it is not precise to assert that the sand layer is reducing landward grain size. Although the laser diffraction method offers high resolution, reliability, and reproductivity (Shimadzu 2012), the number of samples limits the quality of the results and increases the expected error for interpretation.
On the other hand, regarding sample preparation, sample homogenization using a spatula after organic matter dissolution may not be enough for sample disaggregation, which could leave some particle aggregates that could mislead the measurement. Thus, to improve the precision of the results, it is desirable to increase the sample volume and the number of sample measurements. This reduces systematic errors.
The sand layers, S5, S6, S7, and S8 exhibit slight or no lateral change in the probabilistic distributions of the grain size. On the other hand, S4, S9, and S10 show a remarkable variation from sand at H02 to mud at H04. Interestingly, although S5 and S8 are the thickest sand layers and exhibit the maximum landward extent, the landward decrease in the mode value is not significant. This can be explained by the homogeneity in the source's grain size or greater flow speed and depth for the distance between H04 and H02, which is only 200 m (Fig. 3). However, some sedimentary characteristics can be observed in the grain-size probabilistic distributions, which fluctuate along with the sand layer showing the particles’ behavior when they settle down. In sand layers, such as S5 in H02 and S8b in H02 from the middle to the upper part, and S8 in H03, in addition to upward fining, kurtosis tends to be reduced, and skewness tends to rise owing to a lower grain selection (Fig. 7). On the other hand, in S4 in H02 and S8 in H02 from S8a up to the middle of S8b, kurtosis and skewness behave oppositely, suggesting upward higher grain segregation. As described previously, some distribution curves exhibit multimodal distribution due to the mixing of sand-mud or sand-granules mixing. Coarser particles exhibiting negative phi values in a secondary mode peak are composed of rounded pumice grains and are distributed mostly towards the base and top of the layer (S5 in H02 and S8 in H02), and further inland (S5 and S8 in H04) (Fig. 7, Table 2). This can be explained by vertical sediment stratification by density during sediment transport in the inundation flow and differences in particle settling velocity.
In the mud layer interbedded with S9, the rip-up clast was derived from the mud layer below S9a. The front (Fig. 4c) and lateral (Fig. 4b) views expose differences in axial lengths, which points out the possible transport direction perpendicular to the a–axis; hence, perpendicular to the coastline. In the CT images from cores H02, H03, and H04 (Fig. 4), units S6, S9, and S10 exhibit a remarkable landward decrease in thickness. In addition to the landward decrease in magnetic susceptibility (Fig. 5), landward thinning reflects the progressive reduction in energy and flow capacity during inundation from the sea (Fig. 4).
The lateral variation of MS in S7 is also different, with a maximum in the middle section of the transect (Fig. 5). As high values of magnetic susceptibility can be associated with high-energy environments (Černý et al. 2016), it is possible to suggest a flow velocity peak reached at that point. Even though the inverse grading is observed in S7, diatom evidence points to its marine origin.
As mentioned before, Velasco et al. (2019) compared the inundation capacity of both storms and tsunamis in the area of this study by means of numerical modeling using the present–day topography without coastal engineered dikes. For the storm calculation, extreme parameters were utilized for a cyclone, which is unrealistic for Hachinohe because the trajectory of the storm was NE to SW (counter–clockwise), opposite to the expected Coriolis–induced SW to NE direction. Even under such extreme boundary conditions, the storm surge could not inundate and deposit sediments on areas higher than 3 m in elevation (Fig. 2a). For the tsunami surge, under the source parameters of the 2011 Tohoku–oki event, the calculated wave was capable of reaching the 5m–uplifted terrace (Fig. 2b). Therefore, closer or more potent tsunami sources are capable of inundating the study area further inland.
As will be described in the environment reconstruction section (see the discussion section in the body text), the sedimentation rate is constant from 5500 cal BP to 2000 cal BP, indicating that the sedimentary environment is associated with tectonic stability (Fig. 15). Hence, large vertical crustal movements did not occur, and the paleosurface in which the S10 event was settled was located approximately 1.7 m under the current topography (Fig. 3). The sand units identified as marine origin (i.e., S4, S5, S6, S7, S8, S9, and S10) were deposited on the exclusive tsunami domain zone, that is, areas above three–meters–high unaffected by the storm’s influence (Fig. 3). Thus, the sand layers can be categorized as a result of the tsunami surges.
5.3 Sedimentation processes
5.3.1 Implications from ITRAX and CT data
The CT profile displayed bulk density changes associated with landward thickness reduction, packing, and grading (Figs. 4 and 7). Likewise, ITRAX–pattern analysis is significant not only for geochemical characterization but also for sedimentological analysis by showing elemental changes related to the clastic composition variation (Fig. 6 and Appendices 3, 4, and 5). Such trend changes in the sand layers exhibit a close relationship between the elemental pattern and changes in grain size, packing, and composition (Figs. 4, 6, and 7 and Appendices 3, 4, and 5). As the mud is mixed into the sand layer, the clastic signal (i.e., sand fraction) decreases proportionally with the sorting reduction and composition change (Fig. 7). Thus, the upward trend of the decrease or increase in Ca and Sr patterns in the sand layers can be associated with the sand-mud ratio change, respectively. As these two elements (Ca and Sr) have similar atomic properties, they can easily replace each other in the lattice (Nichols 2009) and can be related to the same origin. In addition, changes in Ti display variations in the content of heavy minerals (Arnaud et al. 2015) within sandy deposits, reflecting the density–gradation by a decrease in the particle settling velocity, due to the landward change in MS. Similarly, Rb is related to the mica content (Mills 1964), and these phyllosilicates have low density and low settling velocity. Its deposition takes place in the last lapse of the flow, increasing upward along with an increase in the mud portion.
Ta has high values in the peaty–mud layers because it tends to be dissolved by weathering and concentrated in the latest phases of fluvial sediment transport, depositing as mud on floodplains (Parker and Fleischer 1968; Fricke and Heilig 2006). Thus, Ca/Ta represents the subtraction of the mud content from the total Ca (i.e., plagioclase), which means that this ratio reflects the relative siliciclastic content along with the sediment core sample and its trend variation. As depicted by the rows in Fig. 6 and Appendices 3 and 4, when the Ca/Ta value decreases, the grain size and packing also decrease (Figs. 4 and 7). Consequently, the Ca/Ta pattern mimicked the relative grainsize and packing change. Ta was also helpful for normalizing the elements used in the LDA. As mentioned above, it is closely related to mud (Fig. 6 and Appendices 3, 4, and 5), and it can be considered relatively homogeneous and abundant in the core samples. Thus, it serves as a reference for the comparison of the elements related to the sand layers.
On the other hand, CT imagery suggested a rapid landward reduction in sand content and density via a reduction in color intensity in units S8 and S5.
5.3.2 Implications from grain-size data
Although the grain size analysis results are not sufficient for inferring a marine origin, it is still helpful to understand the sedimentary process. The inverse grading appeared in S7 and at the base of S8, both in H02 (Fig. 7). Three different mechanisms can explain this phenomenon. Firstly, Sohn (1997) described the traction-carpet model as a result of a bivariant velocity profile, which results in upward coarsening along with an increase in velocity. Such a model has been developed based on turbidites and hyperconcentrated flow deposits (Sohn 1997), as well as previous paleotsunami studies (e.g., Moore et al. 2011; Minoura et al. 2013). Secondly, Yoshii et al. (2017) proposed that the inverse grading at the base of the deposit may be related to processes such as kinetic sieving, geometrical sieving, and spatial differences in flow speed (e.g., Middleton 1970; Kneller and Branney 1995; Dasgupta and Manna 2011), rather than traction- carpet sedimentation. Thirdly, Naruse et al. (2012) assumed velocity changes with time, including acceleration and deceleration cycles, which induce upward coarsening and fining, respectively, and are divided by an internal erosion surface (IES). The IES can be defined as the sedimentary truncation generated by the maximum shear velocity.
We interpreted the boundary between S8a and S8b as an IES (Fig. 4). Thus, both S8a and S8b corresponded to the same run-up event and represent a flow cycle of acceleration–deceleration, respectively. The presence of the two subunits in H02 can be explained by the reduction in internal sedimentary units along with landward thinning (Naruse et al. 2012).
All three explanations fit the inverse grading sedimentation style. Although Sohn’s model employs a logarithmic profile of flow velocity and its increase with depth reduction and Naruse’s model assumes velocity and acceleration changes with time, Yoshii’s model includes geometrical properties. One does not necessarily exclude the other. In that sense, during the acceleration phase, a traction-carpet sedimentation style can be developed to generate inverse grading, such as in S8a. Undoubtedly, it is necessary to deepen the understanding of inverse-grading deposits in tsunami events.
Based on the sedimentological characteristics of the sand layers, it is possible to classify them into three types of sedimentation: (1) normal grading with fast landward sediment run–out, marked by a strong landward grain size change from sand to mud and landward thinning (S4, S9, and S10); (2) massive sands with slight vertical and lateral changes (S6 and S7), and (3) normal grading with thicker and broader sediment deposition (S5 and S8). Such settling can be related to three aspects: first, the sediment–source distribution and availability, which vary seasonally in the shoreface (Dean and Dalrymple 2004); second, the paleosurface formed by successive ponds that cause a rapid run out of the sediment by a rapid decay in the sediment transport capacity of the wave; and third, the surge’s energy is controlled by the source magnitude and distance.
5.5 Recurrence interval and regional correlation of tsunami events
The calibrated 14C results exhibited an average recurrence interval of 545 to 1240 years (2σ) for all sand layers (Table 5), during the period between 1200 and 5500 yr BP. However, the lack of marine evidence indicates that sand layers S1, S2, and S3 cannot be directly correlated to tsunami events, and their inclusion in the recurrence estimation reduces the reliability of the calculation. Thus, a recurrence interval of 545 to 736 years (2σ) was obtained from S4 to S10, from ~2700 to ~5500 yr BP. This does not imply that no tsunami affected the Hachinohe area between ~1200 and ~2700 yr BP. Instead, it suggests that there was no tsunami evidence recorded or the geological conditions for such time were different and left tsunami evidence without a clear marine imprint.
On the coast of Misawa city, Tanigawa et al. (2014) identified a tsunami deposit formed from 2900 to 4800 cal BP, which can be broadly correlated with events S4, S5, S6, S7, S8, and S9 in the present study (Fig. 15). To the south, tsunami events from Hachinohe can be correlated with events found by Takada et al. (2016) in Harashinai in Hirono Town (Fig. 15). Such chronological correlations lack sufficient accuracy owing to the error ranges of the calibrated 14C ages.
Based on an exclusive chronological correlation, the recurrence interval inferred from paleotsunami deposits in Hachinohe is slightly shorter than that estimated in Harashinai, which suggests an average of 530 to 956 years (2σ) for the same time frame (~2700 to ~5500 yr BP; Fig. 15). The greater interval in Harashinai can be related to two factors: first, the confidence intervals (2σ) of the probabilistic ages for TSb8, TSb9, and TSb10 are considerably broad, which causes an equally broad probabilistic distribution for the calculated recurrence; second, Harashinai and Hachinohe face both trenches at different angles and at different latitudes. This means that evidence from Hachinohe can be related to the activity of the Kuril and Japan trenches and the their junction, while the evidence recorded in Harashinai can be primarily related to the activity of the Japan Trench (e.g., Figure 7 in Tetsuka et al., 2020). Such geological records and recurrence intervals reaffirm the importance of Hachinohe as a suitable site for paleotsunami research.
To improve temporal correlations and recurrence estimation, it is imperative to increase the precision of the deposition ages by conducting detailed studies at every site. Such studies must include rip–up clast dating and high–resolution systematic dating, coupled with a Bayesian analysis (Ishizawa et al. 2017; Ishizawa et al. 2018; Ishizawa et al. 2020). Complementarily, the application of optically stimulated luminescence (OSL) can contribute to constraining deposition ages and as a complementary source–discrimination proxy (López et al. 2018). In addition, numerical modeling of the tsunami sources will improve our understanding of the size of tsunami events that are capable of inundating the tsunami deposit sites examined here (e.g., Figure 7 in Tetsuka et al., 2020).