[H-DS05] Landslides and related phenomena

[HDS05-03] Grain Size Distribution Characteristics of Debris flow Torent in Ohya Landslide, Central Japan

*Samikshya Dahal¹, Imaizumi Fumitoshi¹ (1.Shizuoka University)

Keywords:Debris flow, Grain size, Automatic object detection

Debris flow is a common hazardous phenomenon that causes several sediment disasters because of its characteristics of high velocity, long travel distance, and destructive power. Debris flow poses a serious threat to people's lives and property, so large effort is needed to investigate and explore different aspects of debris flow in debris flow torrents. The aim of this study is to the automatic measurement of grain size distributions of boulder size particles of the debris flow torrent in the Ohya landslide, central Japan, to determine the sediment deposition characteristics at channel deposits of the debris flow torrent. In this study, we have investigated grain size distributions by combined use of BASEGRAIN software and aerial photographs taken from remote controlled UAV. This study involves mainly two steps; UAV-based Structure from motion (UAV-SfM) photogrammetry of debris flow torrent of Ohya landslide and granulometric analysis of those photographs using object detection software BASEGRAIN. We have used multiple clipped images that are exported from orthomosaic generated by UAV-SfM. By utilizing a sophisticated 5-step object detection algorithm, BASEGRAIN assigns the a-axis and b-axis to each detectable grain. In our results, only large-size surface sediments with a long axis bigger than 15 cm could be detected because the UAV images were obtained from a greater distance of around greater than 50 m height. To fully rely on analysis from BASEGRAIN, the ideal environment must be free of intergranular noise and vegetation, uniform lighting, and dryness. We performed many visual tests on each image to remove errors that were brought on by poor image quality and those unfavourable conditions. After comparing the BASEGRAIN analysis data obtained from different sections of channel deposit, it is clarified that the grain size of channel deposits changes across upper and lower channel reaches, also grain size varies with time even in the same channel section following debris flow events. This study implies that the information gathered on grain size will be useful in the future for estimating other parameters, such as the velocity, energy, and volume of debris flows, which can be used to support the development of early warning systems as well as hazard mitigation and risk reduction plans.