Estimation of design rainfalls considering nonstationarity in observations

Abstract

It is observed that annual maximum rainfall series have an increasing trend in Korea, which may be affected by recent heavy rainfalls. This study proposes appropriate methodology to estimate design rainfalls considering the observed increasing trend in annual maximum rainfall data. The GEV distribution is used in this study. The GEV distribution has three parameters: location (Plo), shape (Psh) and scale (Psc) parameter. This study focuses on estimating the parameters considering the recent trend of them. We first estimate the location parameter based on the empirical relationship among the mean rainfall, and the recent mean rainfall, and the location parameter. And, then, we estimate the shape and the scale parameter based on the regression equations. Two regression equations ( , ) are derived from re-sampling data using the bootstrapping technique. Finally, we estimate the probability density function (PDF) of annual maximum rainfall based on the parameters of the GEV distribution. The PDF is the nonstationary distribution which represents the recent trend in observed data. Using hourly rainfall data collected for 46 years at the Seoul rainfall gauge station in Korea, we estimate the PDFs in condition of stationarity and nonstationarity. We discuss the usefulness of the nonstationary frequency analysis in hydrological planning and designing in changing environment.