Probable precipitation is estimated from a complete-duration data set of the AMeDAS data, which has a dense station coverage over Japan, instead of annual maximum precipitation data sets used in previous studies. Thus, we can obtain probable precipitation with a relatively small estimation error due to the large size of the data set. The probable precipitation is estimated by applying the gamma distribution to the AMeDAS data. The results are summarized as follows:
1) The geographical distribution of probable daily precipitation is closely related to topography, and that of probable hourly precipitation shows large latitudinal difference. In Kyushu, the high-ranked stations in probable hourly precipitation are distributed in the western part, and those in probable daily precipitation are found along the backbone range. Also, in eastern Shikoku, the probable hourly precipitation and probable daily precipitation are greater in coastal regions and mountainous regions, respectively. These differences indicate that the duration and intensity of heavy rain differ with geographical factors, such as topographical conditions and latitude, its cause having peculiar regional characteristics.
2) The relationship between probable precipitation R(t) and time t can be expressed by the equation R(t)=ct
n. A station with a large value of n has long duration of heavy rains. The value of n at each station changes slightly with the length of a return period, and the regional difference in the value becomes increasingly large with an increase in the length.
View full abstract