Abstract
We produced a preliminary high-resolution probable hourly precipitation (mm h−1) and probable Soil Water Index (SWI) for a 50-yr recurrence interval over the Japanese archipelago from 5-km grid-cell Radar/Raingauge-Analyzed Precipitation (R/A) with a 26-yr time series (1988-2013). To date, the selection and applicability of a probability density function such as the Gumbel distribution (GUM) or the generalized extreme value distribution (GEV) has been investigated mainly from a statistical perspective, whereas this study examines them from the viewpoint of disaster prevention. Results show detailed spatial patterns of the probable hourly precipitation and SWI in Japan with ranges of 17.0-158.0 (±1.1-51.1) mm h−1 and 82.1-638.6 (±3.7-93.0), respectively, by spatial resolution of 5-km grid-cell. The probable hourly precipitation by GUM is less than that of GEV with a shape parameter of less than 0.0. Our results demonstrate that R/A-based probable hourly precipitation and SWI by GUM, which has robust estimation, is appropriate for landslide hazard assessment. There results revealed the potential and limitation of estimating the probable precipitation from the R/A dataset. Further study should verify the R/A dataset and the probable precipitations with accumulation of R/A.
