[M-IS23_2AM1] 津波堆積物

[MIS23-02] 機械学習を用いた津波堆積物の地球化学的判別

*桑谷 立¹、永田 賢二²、岡田 真人²、渡邊 隆弘¹、小川 泰正¹、駒井 武¹、土屋 範芳¹ (1.東北大学大学院環境科学研究科、2.東京大学大学院新領域創成科学研究科)

キーワード：津波堆積物, 機械学習, 地球化学

Tsunami deposit is a direct evidence of inundation area of past tsunamis. A large number of publications have been written about the diagnostic signatures and identification criteria for past tsunamis, including sedimentological, micropalaeontological evidences. However their identification is still difficult because all criteria is neither necessary condition nor sufficient condition due to various origin, mechanism and temporal variation of tsunami deposits. Geochemical discrimination is now recognized as other useful proxy which dose not depend on the researcher's subjectivity, especially in the case that other proxies can not be used. Especially, geochemical indicator is suggested to be useful in identification beyond the limit of recognizable sand deposition. In this study, we established the criteria for geochemical discrimination of 2011 Tohoku-oki tsunami deposits and their background marine sediments using machine learning techniques. For 18 analyzed elements, several tens of elemental combinations show the discrimination rates higher than 99%. By applying the criteria to past tsunami deposits in the Sendai Plain, we discuss the validity and effectiveness of the method.