(c) Consider whether an adjustment to the probability assigned to a high outlier data value can be

justified. Often a high outlier is due to a storm event and extreme hydrodynamic response which are much

more severe than would normally be expected over the length of hydrodynamic record. Meteorological

records generally cover a much longer historical time period than hydrodynamic records. By carefully

analyzing storm probabilities and longer-term records from nearby sites if available, it may be possible to

assign a more realistic (lower) probability to the hydrodynamic outlier. Then a more valid distribution

function fit can be obtained.

(5) Choosing an extremal distribution function. When several candidate distribution functions are under

consideration, usually one is selected as a best fit to the data. The selection criteria can range from visual

inspection of plotted results and simple statistics such as the correlation between data and model (e.g.,

Leenknecht, Szuwalski, and Sherlock 1992) to more elaborate statistical tests (Mathiesen et al. 1994). An

objective approach to selecting a distribution function for significant wave heights is given by Goda and

Kobune (1990).

(6) Confidence intervals. Confidence intervals associated with the chosen distribution function should

also be estimated, preferably with a computer program (Leenknecht, Szuwalski, and Sherlock 1992; Goda

1988; Goda 1990; Mathiesen et al. 1994). They depend on the distribution function and number of data

values. Confidence in computed values can also be influenced by random and systematic errors in the data

and physical site characteristics such as long-term variability of water level and climate, possible extreme

events not represented in the recorded population, and physical limits on extremes (such as the depth-imposed

limit on wave height in shallow water).

(1) Extreme conditions in coastal engineering are often described in terms of *return values *and *return*

equalled or exceeded. For example, a 25-year significant wave height is that height that is equalled or

exceeded an average of once during a 25-year time period. Return period is expressed as

(II-8-1)

^

1 & *P *(*H*s)

where

(2) A related concept, *encounter probability*, gives the probability that waves with *H*s equal to or greater

than *H * s will occur during the design life or other time period. It is given by

(II-8-2)

II-8-12

Hydrodynamic Analysis and Design Conditions

Integrated Publishing, Inc. |