Based on an old idea of Leon Borgman, we discuss a unified approach to extreme value analysis of ocean waves. It is shown that extreme wave and crest heights, as well as the significant wave height can be analyzed using similar methods. The methodology can be combined with bootstrap methods for assessing the reliability of the estimates. General implications of the methodology are summarized, and applications to various real and simulated data sets illustrates its use.
Estimation of extreme wave heights, significant as well as individual wave or crest heights, is an important part of many metocean studies, and numerous methods have been developed (Tucker, 1981; Goda et al., 1994; van Vledder et al., 1994; Regnault et al., 1994; and the references therein). In the present paper we shall elaborate on an old idea by Leon Borgman (19731 on how to compute the maximum wave height occurring in a hurricane. It turns out that Borgma~fs idea can be extended to include not only wave height, but also crest height and significant wave height. Moreover, the method ca~l be combined with bootstrap tectmiques for assessing the reliability of the estimates. The amount of wave data will always be limited when considering design for time spans from 50 to 100 years, and the key points are (i) to make optimal use of the available data, and (ii) to know the reliability of the obtained estimates. The main tool will be to look at the wave conditions as nested, or multiple level stochastic models (Athanassoulis et al., 1992), and in fact, there are a series of time scales involved. The individual wave period is O(10 seconds), the temporal correlation of the surface elevation at a fixed location 0(2 minutes), and the duration of a fairly stationary sea state (.9(1 hour).