Extreme water level analysis at three stations on the coast of the Northwestern Pacific Ocean

Abstract

In this study, the data from three long-term observation stations, Aburatsu, Xiamen, and Hong Kong, which are located on the northwest Pacific Ocean coast, were analyzed to estimate the 100-year annual maximum water levels. The performances of four common frequency analysis methods, namely the Gumbel, Weibull, GEV, and GPD distributions, were evaluated. It is found that the GEV model performs best among these four distribution models in Hong Kong and Aburatsu, whereas the Gumbel distribution is the best at the Xiamen station. It is also found that the GEV model generally performs better than the Gumbel model in regard to the mean high correlation coefficient and the mean minimum root-mean-square error. Moreover, in this study, the r-largest value model was used to study temporal trends in the 50-year annual maximum water levels on the northwest Pacific coast over the past fifty years using the observation data of Hong Kong, Xiamen, and Aburatsu. The results show that there are two temporal features in the 50-year return levels at all three stations, with the first being an overall increasing trend over the whole period and the other being an oscillatory trend over the period of observation. The relationships between the temporal trends and the Pacific Decadal Oscillation (PDO), sea level rise, and change of typhoons were also analyzed in this paper. It is found that when the PDO index is shifted to be 4 years in advance, a significantly negative correlation will occur between the PDO index and the 50-year return levels. However, sea level rise and changes of typhoons cause the overall increase over the entire period.