In this study, frequency analysis based on the Peak-Over-Threshold (POT) method was performed to reflect the non-stationarity of extreme sea level. Of the parameters of the general Pareto distribution (GPD), the scale parameter was considered as a time-dependent function. The Busan site was selected as a study site, and sea level data from 1976 to 2015 (40 years) were collected at 1-hour intervals. The threshold value of the POT time series was determined to obtain extreme sea level estimates as close to those calculated by the generalized extreme value distribution based on the annual maximum time series. By comparing the stationary GPD model with the non-stationary GPD model, it was determined that the estimate of the extreme sea level for the 50-year return period is likely to be increased by approximately 83 cm in 2055.

In Republic of Korea, stationary extreme value distribution models are generally used for estimating the design of coastal and harbor structures. However, due to the impact of climate change, the probability of typhoon is recently increasing. In order to consider this tendency, a non-stationary extreme value distribution model was applied in this study. The annual maximum storm surges were calculated by the storm surge model, and then, results from the storm surge model were applied to a non-stationary GEV (Generalized Extreme Value) distribution model to calculate the extreme storm surge height. The storm surge height of 50 year return period by non-stationary model was 28% higher than the storm surge height of a previous study that used stationary models. The overall results achieved in this study are expected to provide data and useful in establishing a master paln for disaster prevention.

No abstract