Stochastic Model for Estimating Extreme Water Level in Port and Coastal Engineering Design

Abstract

Extreme water level is an important consideration when designing coastal protection structures. However, frequency analysis recommended by standard codes only considers the annual maximum water level, whereas water levels should actually be regarded as a combination of astronomical tide and storm surge. The two impacting factors are both random variables, and this paper discusses their dependency structures and proposes a new joint probability method to determine extreme design water levels. The lognormal, Gumbel, Weibull, Pearson type 3, traditional maximum entropy, and modified maximum entropy distributions are applied to fit univariate data of astronomical tides and storm surges separately, and the bivariate normal, Gumbel-Hougaard, Frank and Clayton copulas are then utilized to construct their joint probability distributions. To ensure that the new design method is suitable for use with typhoon data, the annual occurrence frequency of typhoon processes is considered and corresponding bivariate compound probability distributions are proposed. Based on maximum water level data obtained from Hengmen hydrological station in the Pearl River Basin, China, these probability models are applied to obtain designs for extreme water levels using the largest sum of the astronomical tide and storm surge obtained under fixed joint return periods. These design values provide an improved approach for determining the necessary height of coastal and offshore structures.