Testing a failure surface prediction and deposit reconstruction method for a landslide cluster that occurred during Typhoon Talas (Japan)

Abstract

Abstract. Reconstructions of failure surfaces (prior to\npotential landslides or after their release), landslide deposits, or other\npalaeotopographic features are important for hazard and erosion assessment.\nThe volumes involved in landslide and failure surfaces constrain the\npropagation of a landslide, and knowledge of the past topography helps\nus to understand these hazards. Some methods exist to characterise landslide geometry, but these methods\nusually require monitoring information. This study tries to assess the\nvalidity of the sloping local base level (SLBL) method for this purpose. Two\nsets of airborne lidar digital elevation models (DEMs) of the Kii Peninsula\n(Japan) are used: the first one was acquired before Typhoon Talas, and the\nsecond one was acquired after. A total of 70 deep-seated landslides occurred\nduring this event between 2 and 5\xa0September\xa02011. This study shows that the SLBL method is efficient using either the slope\ndeformations identifiable on the DEM before the release of the landslide or\na reliable 2.5-D failure surface created by using both DEMs (the 2.5-D\ncorresponds to a surface which has only one z value for each\n x – y coordinate; in other words, no true vertical topography or overhang can be\nrepresented perfectly). In addition, this method allows for the reconstruction\nof eroded deposits and buried valleys. Most of the volumes estimated are\nwithin ±35 \u2009% of the estimation made by Chigira et al.\xa0(2013), and\nthe coefficients of expansion range from 10\u2009% to 25\u2009%. These results show\nconsiderable sensitivity to the parameters used for the reconstruction\nof the landslide volume estimations and demonstrate the need for an\nefficient and fast tool to reconstruct potential landslide geometries or\nhistories.