抄録
This paper describes the applicability of probability distributions with both upper- and lower-bounds to hydrologic frequency analysis. Two distributions are applied to some data sets of extreme-value precipitation and river discharge: the Slade-type four-parameter log-normal distribution and the extremevalue distribution with upper- and lower-bounds (EVLUB) proposed by Kanda (1981). Their goodness of fit to the data sets is assessed in terms of the standard least-square criterion (SLSC). The analysis using the bootstrap method indicates that these distributions with the upper-bound give less variability in quantile estimates than the three-parameter log-normal distribution with the infinite upper-bound. Finally, this paper discusses incorporation of the probable maximum precipitation (PMP) or probable maximum floods (PMF) into the frequency analysis models as the upper bounds.
