View of tsunami deposit outcrop at Jyuosaki, Noda Village, Iwate Prefecture, Japan. Photo at the upper left is the cover photo of Part I, taken by a drone from offshore (white circle in the photo indicates the location of the outcrop in the middle photo). The photo at the right is a close-up view of the trench indicated by the arrow in the middle photo, where several tsunami deposits can be identified; the sand layer below the To-a Formation (Towada a tephra, A.D. 915) is presumed to be the Jogan tsunami deposit (A.D. 869). Traces of paleo-tsunamis occurring during the last 6500 years can be observed in the whole outcrop, which is over 7 m in height.

(Photograph & Explanation: Noriyoshi TSUCHIYA and Kengo NAKAMURA)

 A method is developed to quantitatively correlate geological layers based on similarities in the shape of the statistical frequency distribution of a large volume of multi-element count data obtained with an X-ray fluorescence (XRF) core scanner. A distance measure between probability distributions called Jensen–Shannon divergence is adopted as a criterion for similarities in statistical distributions with the assumption of a Gaussian distribution. Using artificially created elemental count data, the flow of analysis and the effectiveness of the method for detecting the query layer of interest from the search target core dataset is demonstrated. By applying the system to geological samples, which were disturbed by the 2011 Tohoku-oki tsunami, located in Higashi Matsushima City, Miyagi Prefecture, the system is shown to appropriately correlate surface layer, Jogan-tsunami (A.D. 869) layer, and beach sediment layer, which indicates the effectiveness of the proposed system for obtaining a stratigraphic correlation of two cores. In the future, by developing the method to automatically determine layer boundaries, it will be possible to detect narrow event layers and to automatically correlate the stratigraphy. By applying it to many cores, the proposed method is useful for evaluating spatial distributions of tsunami deposits and wide-spread tephra layers, and it is expected to contribute to disaster prevention and mitigation.

 In order to establish a discrimination method for tsunami deposits, a machine learning analysis is conducted for geochemical data to determine paleo-tsunami deposits. Column samples containing tsunami deposits are collected at Noda village, Iwate prefecture, and Wakabayashi-ku, Sendai city, and the distribution of element concentrations are continuously measured. The model is trained by Multilayer perceptron using Noda samples as training data. Combination of elements and number of layers and perceptron are determined by the brute-force search method applied to the Noda samples. The results show that all event deposits determined in the Wakabayashi samples are tsunami deposits. These results indicate the possibility of highly accurate discrimination without being affected by sampling points or depositional ages, or by selecting appropriate supervised data. To combine the techniques of machine learning and geochemical discrimination, simple determination systematics are developed for tsunami deposits using supervised data and analyses of evaluation data.

 From the perspective of disaster prevention, it is necessary to clarify the distribution and historical frequency of tsunami deposits. However, tsunami deposits are disturbed by earthquakes and tsunamis. Therefore, it is difficult to identify the layers of deposits. Layers deposited during disturbances are dated and grain-size distribution is measured. Using the results of geochemical data and principal component analyses with detailed elemental distributions, stratigraphic boundaries of past tsunami deposit layers and other layers left in the disturbed deposits are estimated. Samples collected with a geo-slicer consist of a 50-60 cm deep disturbed layer deposited by the Great East Japan Earthquake and underlying sand, peat, and marine deposits. 24 elemental distributions in a core measured using Itrax indicate that the disturbed layer is dominated by heavy metals. Changes in principal component analysis scores infer traces of a layer that could not be visually identified in the disturbed layer. Traces of this layer are consistent with traces that are deposits indicated from the isotopic analysis and the grain size distribution. Therefore, even if it is difficult to identify deposits due to disturbances caused by the tsunami, it may be possible to identify unidentified layers and estimate their thicknesses by compressing their dimensions and clarifying elemental relationships.

 To improve a method for geochemically discriminating paleotsunami deposits, a quantitative analysis of major and trace elements in boring cores from the Shizuoka plain on the Pacific coast of central Japan was performed using a portable energy dispersive X-ray fluorescence spectrometry system (portable XRF). Geochemical approaches using a portable XRF contribute to studies on simple and rapid methods for detecting paleotsunami deposits. Calibration curves for quantitative analyses were prepared using 25 geochemical standard samples. Geochemical characteristics of paleotsunami deposits (ca. 1000, 3500, and 4000 cal BP) from the Shizuoka plain were obtained. Most of the quantitative data collected with the portable XRF were good agreement with reported values measured with stationary-type energy dispersive XRF in previous studies. Therefore, a portable XRF was applied in geochemical studies of samples from the Shizuoka plain. Relatively high Si/Ti, Ca/Ti, and Sr/Ti atomic ratios were observed in paleotsunami deposits from layers aged ca. 1000, 3500, and 4000 cal BP. Moreover, relatively high Cr/Ti ratios were found in lower paleotsunami deposits (ca. 4000 cal BP). Fe/Ti and S/Ti ratios increased in peaty clay layers just below the paleotsunami deposits in the core. On the other hand, Cl was not detected in the cores. Most of the paleotsunami deposits were distinguished from other layers in the cores using a cluster analysis with the portable XRF data. Therefore, a portable XRF is useful for characterizing paleotsunami deposits from the Shizuoka plain.

 Suruga Bay is a tectonic bay that is traversed in the north-south direction by the Suruga Trough, which is considered to be a subduction boundary between the northern margin of the Philippine Sea Plate and the Eurasian Plate. In order to clarify the shape of the subducting plate boundary, seismic refraction surveys were conducted using 14 sets of pop-up type ocean bottom seismographs (OBSs) and air-guns as controlling seismic sources at a total of six survey lines in the years 2016, 2017, and 2018. The main survey lines are located along the western side, the axis of the Suruga Trough, and the Izu peninsula side of Suruga Bay; the three main lines run approximately north to south across Suruga Bay. This enables the subsurface velocity structure beneath the sea floor to be obtained. Features indicate the shape and subduction of the subducting plate boundary if the subsurface velocity structure beneath each survey line can be determined in detail. Details of the velocity structure at the western side of Suruga Bay investigated in 2016 are presented. The survey line 2016-NS is located at the Eurasian Plate side and runs about 36 km from the Senoumi Basin to off the Udo Hills at the west side of the Senoumi-kita Bank, almost parallel to the west coast of Suruga Bay. Survey line 2016-EW crosses Suruga Bay about 18 km from near Yaizu on the west coast of the bay to near Toi on the east coast. The subsurface velocity boundary is complex, and an unusual structure rises in the subsurface structure at the north side of Senoumi-kita Bank. The rising structure is connected to the Senoumi-minami Bank and the Senoumi-kita Bank, and is considered to be an uplift derived from the outer ridge uplift belt zone.