Abstract
Future changes of probability distributions of extreme rainfall and flood are analyzed using “Database for Policy Decision-Making for Future Climate Change, d4PDF” having 60-years climate simulations with 50 ensemble members for the historical experiment and 90 members for the future experiment. The magnitude of a largest-class flood equivalent to a 1000-years flood is also analyzed. The study basins are the Ara River basin (2940km2), the Shonai River basin (1010km2), and the Yodo River basin (8240km2) in Japan. The results show that 1) the frequency distributions of annual maximum 24-hours rainfall for d4PDF of the historical experiment match with the ones for observed data in the range of the non-exceedance probability less than 0.95; 2) the 200-years annual maximum 24-hours rainfall for the d4PDF of the future experiment is 1.3 to 1.4 times larger than the one of the historical experiment and 1.5 to 1.7 times larger for the annual maximum river discharge; 3) the 200-years annual maximum 24 hours rainfall for the d4PDF of the future experiment is equivalent to the one of 900 years of the historical experiment; and 4) the rainfall patterns that cause largest-class floods simulated by d4PDF well match with the one of large historic floods for the three basins.
