In this article, we analyzed cases of debris flow caused by shallow landslides in Hiroshima and deep-seated landslides in the southern part of Nara Prefecture and made various considerations on the parameters of the three-tier tank model used for forecasting sediment-related disasters during heavy rainfall. The results showed that although the peak time of the soil water index nearly matched the debris flow occurrence time, there was a considerable discrepancy with the occurrence time of the deep-seated landslide. However, when six of the parameters used in the soil water index calculation were optimized, it was found that the peak time value of the third tank could approach that of the deep-seated landslide onset time. Therefore, the findings suggested that the third tank value may effectively predict the occurrence of a deep-seated landslide during heavy rainfall.

  The 1993 Hokkaido Nansei-oki Earthquake induced many landslides on Okushiri Island and the Japan Sea coast of Hokkaido Island. In this study, GIS-based feature analysis and random forest-based factor analysis were conducted to determine the factors contributing to the landslides induced by this earthquake. The results show that the most important factor contributing to the occurrence of landslides due to this earthquake is the slope aspect, because of the influence of inertial forces at the cessation of ground motion. Finally, this study demonstrates the effectiveness of random forests in assessing the influence of landslides initiation factors.