which represent the intensity of the individual storms, 



64. The availability of synoptic hindcast data base of wave data for 

 most of the US coastline (Corson et al. 1981) accommodates the technique for 

 formulation of F(H) where only the significant wave height Hg values 

 (representing the intensity of a severe storm) above a threshold value are 

 addressed (Battjes 1984). Recent applications of hindcast wave data at WES 

 (Andrew, Smith, and McKee 1985) have yielded good results with a cpd function 

 for significant wave heights above a threshold using the following extremal 

 (Fisher-Tippet) Type I distribution: 



[(e-H )/<J,] 



-e (19) 



F H = e ^^' 



s 



dF(H^) F(H ) r(£-H^)/<t)"| 



s 



where F(Hg) is the cumulative probability that a significant wave height Hg 

 in a sample is equal to or less than some specified Hg , or P[Hg < Hg] . e 

 and (J) are parameters fit to the data by regression. The traditional return 

 period RT can be estimated as (Borgman and Resio 1982) 



RT = 



X[1 - F(Hg)] (21) 



65. Another commonly applied cpd, traditionally used for annual ex- 

 tremes, is the following Weibull distribution: 



IC 



[(e-H^)/*! 



F(Hg) = 1 - e" " -■ (22) 



where C is an additional empirical parameter which must be fit to the data. 

 This distribution is equivalent to a Rayleigh distribution when C = 2 and 

 reduces to an exponential distribution when C = 1 and e = (Petraukas 

 and Aagaard 1970) . 



66. Either of these cpd functions could be applied to estimate the 

 expected damages, given a damage function such as Equation 14. These cpd 

 functions are typically applied to present the probability of exceedance 

 for a specified H (i.e. P(H > H )). Assignment of a representative 

 unit price for repair of displaced armor units allows the expected cost of 



42 



