It is necessary to establish a better method for predicting rainfall-induced debris movements based on current rainfall conditions in order to mitigate sediment-related disasters. For this purpose, a scientific and rational rainfall index that considers regional geological and topographical characteristics should be established. In this study, the author has proposed a warning system for preventing sediment-related disasters that considers regional rainfall characteristics.
In the 1st chapter, the author describes several previous and current disaster prediction methods and their advantages and disadvantages. Although it is extremely important to refl ect regional characteristics for disaster prediction, many problems continue to exist.
In the 2nd chapter, a new rainfall index, denoted as R', is proposed for warning against sediment-related disasters. It is possible to express the influences of both previous and current rainfalls using a single value of R'. The risk can be determined easily using the spatial distribution of R'. To calculate R' with the concept of effective rainfall, a combination of two half-value periods (72 h and 1.5 h) is applied. In the decomposed granite area around Hiroshima Prefecture, the critical values of R' for debris fl ows, hillside landslides, and slope failures were found to be 250 mm, 175 mm, and 125 mm, respectively. The scale of a sediment-related disaster can be indicated by the value of R'.
In the 3rd chapter, the physical characteristics of R' were confirmed by monitoring the relationship between soil water content and rainfall. In general, the slope stabilization problem against rainfall is explained by the occurrence of pore water pressure (due to an increase in underground water levels) and the decrease in shear strength (due to a decrease in suction). During in-situ monitoring, it was observed that the water content of the surface layer increased rapidly even with light rainfall, and that of the basal layer tended to increase with rainfall accumulation. On the other hand, the water content of the intermediate layer increased only when both of the previous rainfall and current rainfall intensity were considerably high. This tendency was also observed in an indoor experiment. In other words, the uptrend of water content differs depending on the rainfall pattern.
In the 4th chapter, the applicability of R' was verified in 8 cases having different characteristics. For example, in Paleozoic rock or rhyolite areas, different coefficients and a larger threshold had to be used for the calculation of R'. Accordingly, the critical value of R' should be varied depending on the local characteristics.
In the 5th chapter, some factors that influenced debris movement were investigated. It was noted that the local characteristics affect the critical value of R' in each region. For example, for evaluating the regional rainfall history, the utilization of the occurrence probability of R' was proposed. As a result, it was shown that the critical value of R' for debris fl ow occurrence in Hiroshima Prefecture could be fi tted with an R' value of 1/8 probability.
In the 6th chapter, the new concept of "remarkable rainfall intensity R_R" is proposed; this index is used for understanding the disaster risk easily, and its usefulness was demonstrated through some examples. R_R was defi ned as "hourly rainfall required for the critical value of R' to become dangerous." For the practical use of R_R in the warning system for disaster prevention, it is necessary for a warning level to be set. R_R can be used in a more intuitive and comprehensible manner in conjunction with the distribution map of R'. The applicability and utility of R_R was confi rmed through case studies in Higashi-Hiroshima City.
The 7th chapter presents the conclusion of this study.
If the current level of risk and the distribution of R' along with information about the remarkable rainfall intensity R_R can be obtained through an early warning system, lives can be saved through the timely evacuation of people before a disaster actually occurs.