Regional Frequency Analysis of Daily Rainfall Observed at Meteorological Observatories in Japan

Regional frequency analysis (RFA) has an advantage in size of usable data over the conventional at-site frequency analysis (SFA). Thus, the RFA is a potential approach for reliable estimation of design rainfall for drainage planning and flood control planning especially in the case that usable data is limited when available record length is short, which is because, for example, observing period is short or length of a climatologically homogeneous period is limited due to long-term climate change. This paper deals with a case study of RFA of daily rainfall observed at 155 hydrometeorological observatories in Japan. In advance to RFA, the observatories were categorized into eight regions by adjusting a result of cluster analysis based on similarity in statistics of daily rainfall and geographical closeness, and validity of the classification was assessed by checking discordancy of each rain gauge in each classified region and homogeneity of each region. The eight regions were similar to region classification for applying flood envelope curve in Japan. The results show that RFA estimates are close to SFA estimates, and bootstrap confidence intervals of RFA estimates were narrower than those of SFA estimates at almost all observatories, which suggests higher confidence of RFA estimates than those of SFA estimates.