of the functions listed in Table 4. The Weibull distribution was computed for 

 values of C = 1.0 and C = 2.0 from Equation 6. These values represent the 

 general range of Weibull shape parameters that are applicable to wave condi- 

 tions. Note that if C = 1.0 , the Weibull reduces to an exponential distri- 

 bution and if C = 2.0 , the Rayleigh distribution (Petrauskas and Aagaard 

 1971). The slope and intercept from Equation 9 are estimated by computing the 

 least squares linear regression line corresponding to Equation 9 (Issacson and 

 Mackenzie 1981). For a listing of estimated values of A and B along with 

 the corresponding parameter estimates for each specific function see Table 5: 



Table 5 

 Estimated Parameters for the Hypothesized Extremal Models 



Model A B ii o 



Extremal Type I 



-6.804 



0.981 



6.936 



1.019 



Lognormal 



-8.058 



1.128 



7.144 



0.886 



Log Extremal 



-14.952 



7.742 



7.742 



6.898 



Weibull, C = 1.0 



-4.615 



0.745 



6.195 



1.342 



Weibull, C = 2.0 



-1.839 



0.363 



5.066 



2.755 



12. Appendix B contains the resulting data plots for each of the pro- 

 posed extremal models. The horizontal axis entitled Cumulative Probability 

 Scale denotes the actual values of the function F(x) from Table 4 that cor- 

 respond to the data values on the vertical axis. The other horizontal axis is 

 the return period as defined in Equation 10. By inspection it is seen that 

 the Lognormal model does not fit. The Weibull with C = 1.0 , or exponential, 

 also displays significant curvature. The Log Extremal model shows less curva- 

 ture, but the plot still deviates from linearity in the upper and lower tails 

 of the distribution. The Extremal Type I and the Weibull with C = 2.0 

 (Rayleigh) both look fairly linear except for the lowest 16 points. Since 

 the choice of 6 m for the data selection threshold was arbitrary, it is in- 

 tuitively appealing to recompute the analysis without the lower 16 points. 

 This results in a total of 62 storm events with maxima greater than or equal 

 to 6.4 m and a revised Poisson parameter of u = 3.1 storms per year. The 

 Chi square goodness of fit test for the Poisson distribution of storms per 

 year was recomputed with the reduced data set. The chi square value was 

 Xh = 1.35 and f Pr Xn ^ 1.35) « 0.85 . The chi square statistic is still 

 smaller than 85 percent of all possible chi square values; therefore, it is 



13 



