concluded that the Poisson model still holds. The extremal plots for the re- 

 duced data sets are displayed in Figures 3 and 4. The Extremal Type I and the 

 Weibull C = 2.0 both appear to fit the largest 62 data points well. The Ex- 

 tremal Type I is the preferred model because (a) it appears to fit better in 

 the upper tail of the data, (b) it is one of the three possible extremal as- 

 ymptotes (Borgman and Resio 1982), of which the Weibull is not a member and, 

 therefore, has a better theoretical basis than does the Weibull, and (c) the 

 Extremal Type I is a two-parameter model whereas the Weibull has three param- 

 eters. The Weibull shape parameter C makes it possible to fit the distribu- 

 tion to data with a higher degree of accuracy than for two-parameter models. 

 This fact does not necessarily mean that extrapolations beyond the data extent 

 will be improved by the higher precision fit. The extrapolations could be 

 very unreliable if the Weibull were used when a simpler model was more appro- 

 priate. Results will be given for both models, but it is believed that the 

 Extremal Type I is qualitatively the better model. Table 6 contains the val- 

 ues of A and B and the resulting distribution parameters. The return 

 periods and associated significant wave heights were computed for both models 

 and ar.e presented in Table 7. 



RETURN PERIOD IN YERRS 



5 10 20 40 60 100 



14.0 



12.0 - 



10.0 - 



8.0 - 



6.0 -4 



1.0 



50.0 80.0 90.0 95.0 98.0 99.0 

 CUMULATIVE PROBABILITY SCALE 



99.9 



Figure 3. Extremal plot for 62 largest maximum 

 significant wave heights. Extremal Type I 



14 



