Narrow Results

In this study, an enormous size sample of typhoon- or low-generated annual maximum (AM) wave height data on the Northwestern Pacific Ocean is produced through a Monte-Carlo simulation over extremely long years, and spatial variation for a parent distribution of storm type-separated or storm-type-free AM wave height data is estimated based on the extreme value analysis and its statistical aggregation. Also each of the results is compared with that for historical storm case. The main findings are as follows: 1) The parent distribution of typhoon- or low-generated AM wave height is well approximated by either the Weibull distribution (typhoon and low cases) or the Gumbel distribution (low case), in case where shape parameter and the other parameters of the parent distributions are space-dependent. 2) Frequent occurrence and passage of strong storms may yield a sharper probability distribution of AM wave height data. 3) The parent distribution of storm type-free AM wave height data is strongly affected by predominant storm type-separated AM wave height data.

A long term wave hindcast system is revised by embedding parametric typhoon model-based wind data into NCEP/NCAR (referred to as NCEP) or ECMWF reanalysis/analysis wind data in order to extend the applicability to the estimation of not only wave climate and its trend but also extreme waves through the improvement of the ability for reproducing typhoon-generated waves. Major findings are as follows. 1) Embedding parametric typhoon winds into the original NCEP wind data yields a significant improvement of availability of the system for the estimation of both wave climate conditions and extreme waves, almost without changing the statistical test results for trend analysis in typhoon-dominated sea areas. 2) Embedding parametric typhoon model winds is less effective in the ECMWF winds than in the NCEP winds, because the ECMWF wind data has a higher accuracy than the NCEP wind data due to its finer space resolution.

Hourly-waves over a period of 51 years from 1948 to 1998 are hindcast at almost all measurement stations deployed on the coastal areas of Japan using a shallow water wave model based on a backward ray tracing method under the input wind conditions extracted from the NCEP/NCAR reanalysis surface wind data archive. Comparison between 2-3 hourly measurements and hindcasts reveals that the system yields reasonable estimates not only for time series of waves and wave climate parameters over a whole measurement period but also for their annual rates of increase or decrease. The important finding is that the 41-year hindcast waves from 1958 to 1998 indicate hardly any statistically significant trend of wave climate on the coastal areas of Japan, whereas the 51-year waves gives a significant increasing trend probably due to the poor quality of wind data in the first 10 years from 1948 to 1957.

Large scale Monte-Carlo simulation is conducted to investigate the accuracy of a method compounding the probability distribution and the variance of return value separately estimated by the extreme value analysis for stratified data. The simulation results suggest that a method compounding separate probability distribution gives fairly accurate estimate of return value in cases where the parameters of each distribution are evaluated by use of a model based on the least square method. Another finding is that the Izumiya theoretical formula for compounding separate variance yields more reasonable estimates compared to the Goda empirical formula. Then the availability of a stratified sampling technique in the extreme value analysis for the measurement data of winds is discussed. It appears that the technique does not necessarily improve the accuracy and the efficiency in an estimate of return value, probably due to augmentation of statistical variability associated with stratified sampling for the measurement data of small sample size. But the method yields more efficient estimates of return value in some cases where typhoon-generated winds are dominant compared to those caused by other types of storms.