Estimating the Parent Distribution of Storm Type-Separated Annual Maximum Wave Height on the Northwestern Pacific Ocean

本文通过对西北太平洋（包括日本海和东海）2万年来台风产生的波浪或低压系统产生的1万年来的波浪进行蒙特卡罗模拟，获得了大样本的风暴类型分离的年最大波高数据，并基于对每个AM波高数据集的极值分析估计了一个母波分布。此外，通过与过去50年历史台风和过去20年历史低点的回波波高及其比值的比较，研究了该方法的有效性。主要研究结果总结如下：各类型历史风暴的50年回归波高分布在定性和定量上都与估算值一致。台风产生的AM资料的母区分布近似Weibull分布，形状参数从西南地区的4左右逐渐减小到西北和东北地区的1.4左右。在日本海，形状参数从日本一侧向大陆方向减小。对于低生成AM数据，在东部地区形状参数从2开始呈半圆形递减，在日本周边广大地区选择Gumbel分布。无风暴型调幅数据的父类分布受到主要的有风暴分离调幅数据的强烈影响。r年回归波高样本与威布尔分布高度近似，且回归周期与样本量之比越大，形状参数值越小。这意味着随着比例的增加，样本分布的尾迹变长。

This study produces a large size sample of storm type-separated annual maximum(AM) wave height data through a Monte-Carlo simulation of typhoon-generated waves over 20,000 years or low pressure system(low)-generated waves over 10,000 years on the Northwestern Pacific Ocean including the Japan Sea and the East China Sea, and estimates a parent distribution based on the extreme value analysis for each of the AM wave height data sets. Also, the availability is investigated through the comparison with the return wave heights and their ratios similarly evaluated for historical typhoons in the past 50 years and for historical lows in the past 20 years. Main results are summarized as follows.1.The distribution of 50-year return wave height is in qualitative and quantitative agreement with that estimated for each type of the historical storms.2.The parent distribution of typhoon-generated AM data is approximated with the Weibull distribution, in which the shape parameter tends to decrease from around 4 on the southwestern area toward around 1.4 on the northwestern area and on the northeastern area. On the Japan Sea, the shape parameter decreases from Japan side toward the continent. For the case of low-generated AM data, the shape parameter decreases semi-circularly from a value of 2 on the eastern area, and the Gumbel distribution is selected on the wide area around Japan.3.The parent distribution of storm type-free AM data is strongly affected by predominant storm-separated AM data.4.A sample of R-year return wave height is highly approximated with the Weibull distribution, and the greater the ratio of return period to sample size is, the smaller the value of shape parameter becomes. This means a longer tailing of the sample distribution with increase of the ratio.