[M-IS07] Geophysical particulate gravity current

[MIS07-P04] Estimation of origin of sedimentary structures in hybrid event beds based on grain fabric analysis: Examples from the Lower Pleistocene Otadai Formation in the Boso Peninsula, Japan

*Ryogo Tanaka¹, Hajime Naruse¹ (1.Graduate School of Science, Kyoto University)

Keywords:Convolutional Neural Network, Semantic Segmentation, Turbidites, Debrites, Flow transformation, Sedimentology

In deep-sea gravity flow deposits, coarse-grained sandstone beds that have both the characteristics of debrites and turbidites are called hybrid event beds (HEBs). HEBs are assumed to contain turbidites in the lower and upper parts and debrites intercalated in the middle part of beds. Thus HEBs are regarded as consequences of flow transformation. However, since most of the sedimentary structures in HEB show massive structures, it was challenging to identify the depositional processes of HEBs by naked-eye observations to verify the flow transformation hypothesis. Here, this study investigates the origin of sedimentary structural divisions of HEBs by field survey and image analysis with microstructures of sedimentary rock cross-sections. This study area is the Pleistocene Otadai Formation distributed in Chiba Prefecture, Japan. As a result, three HEBs are recognized in the succession near the O7 volcanic ash layer in the Otadai Formation. These HEBs are divided into the following five sedimentary structural divisions: Division I (massive sandstone), Division II (clast-supported large mud clasts), Division III (massive sandstone containing mud clasts), Division IV (graded sandstone containing flaky muddy clasts), and Division V (laminated sandstone). Based on outcrop observations, Divisions IV and V showing distinct lamination and graded structures are interpreted as turbidites. Division II, which contains very large mud clasts and is poorly sorted, is interpreted as debrites. On the other hand, grain fabric analysis was needed to understand the depositional process of the massive sandstones (Division I and III). A high-resolution image scanner took images of polished cross-sections of sandstone samples for this purpose. We constructed a convolutional neural network that automatically recognized grains from the cross-section images and measured the grain size and grain fabric of massive sandstones. As a result, the grain fabric of Division I showed the characteristics of typical turbidites, while two types of fabric were observed in Division III imbrication. While the grain fabric of Division IIIb exhibited typical features of turbidites, a high-angle imbrication (40–60°) of Division IIIa indicated deposition from debris flows.
The results of this study suggest that the two sedimentary structural divisions (Division II and IIIa) developed in the middle part of HEB in this study area are both debris-flow deposits. The reason for the transition from poorly sorted debrites (Division II) to well-sorted sandy debrites (Division IIIa) could be considered as a result of two different processes: flow transformation from debris flow to turbidity currents by ambient water entrainment, and that from turbidity currents to debris flows due to entrainment of mud clasts.