Estimation of Secular Change in Return Level of Daily Rainfall Using Metastatistical Extreme Value Distribution

Abstract

The annual maximum series (AMS) method is generally applied to estimate the return level of extreme rainfall. The sample size, however, is so small that the estimated return level and return period are strongly fluctuated even when just one or a few extreme-size minima or maxima are included by change in a target duration for the analysis. That often disturbs correct estimation of secular change in return level and return period of rainfall. In this study, we estimated the secular change in the return level and that in return period of daily rainfall in Japan by applying the metastatistical extreme value (MEV) distribution to daily rainfall data. The results showed that the estimated 100-year daily rainfall by applying the MEV distribution showed similarly increasing trend nationwide as when the annual maximum method was applied. In addition, compared with the AMS method, the fluctuation range of the estimated values was suppressed, and the trend of annual change was shown more clearly.