Application of mineralogy and geochemistry to identify the landward extent of paleo-tsunami deposits not visible to the naked eye in the coastal wetland of Erimo, Hokkaido, Japan

Abstract

The inundation extent of paleo-tsunami can be reconstructed using deposits in geologic records. However, the distribution of visible tsunami deposits does not cover the entire extent of past tsunami inundation. A method for the identification of inundation traces that are macroscopically invisible is required for the accurate reconstruction of the extent of paleo-tsunami inundation. Here, we propose a composite method to recognize tsunami deposits based on mineral particles with high preservation potential. This method is based on multiple proxies, including non-destructive methods such as X-ray computed tomography and X-ray fluorescence scans, and mineralogical and grain size analyses. The method was applied to the paleo-tsunami deposits distributed in Erimo, southern Hokkaido, Japan. Scanning electron microscopy and electron-probe micro-analyzer mapping enable the detection of macroscopically invisible particles. A detailed chemical comparison with the host rocks of the hinterlands revealed that the macroscopically invisible particles share a common source with observable tsunami sand. Their grain size distribution measured using image analysis was compared with that of the observable tsunami sand layer. To verify the inundation area suggested by the macroscopically invisible tsunami sand, we predicted the inundation extent of the paleo-tsunami using the Framework of Inversion of Tsunami deposits considering Transport of Nonuniform Unsteady Suspension and Sediment entrainment Deep Neural Network (FITTNUSS-DNN), a deep-learning inverse model that reconstructs tsunami characteristics from the features of visible tsunami deposits. The maximum extent of inundation predicted using FITTNUSS-DNN was comparable to the distribution of the macroscopically invisible sand, indicating that detecting the invisible sand layer can effectively estimate the inundation extent by paleo-tsunamis.