Spectra were grouped by 30-centimeter intervals of significant 

 height and by variable intervals of peak period (Table 2). Variable 

 period intervals were needed because energy is computed for nonuniform 

 period bands (but uniform frequency bands) in the CERC spectral analysis 

 program. Selection of longer period intervals was necessary so that 

 each interval encompasses only one spectral band. If the intervals had 

 included more than one spectral band, the averaging procedure would tend 

 to produce an unrepresentatively low and broad peak in the average 

 spectrum. 



Table 2. Peak period grouping intervals. 



Period interval 



Frequency bands 



Period interval 



Frequency bands 





in interval 





in interval 



(s) 



(No.) 



(s) 



CNo.) 



1 to 2 



47 



7.8 to 8.5 





2 to 3 



15 



8.5 to 9.3 





3 to 4 



8 



9.3 to 10.3 





4 to S 



4 



10.3 to 11.6 





5 to 6 



4 



11.6 to 13.3 





6 to 6.6 



1 



13.3 to 15.5 





6.6 to 7.2 



1 



15.5 to 18.6 





7.2 to 7.8 



1 



18.6 to 23.3 





Period intervals of 1 second were used for periods shorter than 6 

 seconds (Table 2). Individual spectra with relatively short peak pe- 

 riods are generally more poorly focused in frequency than spectra with 

 long peak periods. Thus, period intervals encompassing more than one 

 spectral band could be used for short periods without creating much 

 artificial broadening of the peak in the average spectrum. This treat- 

 ment is further defensible on the grounds that (a) short -period waves 

 are in relatively deeper water than long-period waves at the measure- 

 ment site and hence exhibit less depth-induced modification, and (b) 

 spectra with peak periods shorter than 6 seconds are generally not im- 

 portant for coastal engineering design applications at ocean sites. 



After the spectra were grouped into height-period intervals, average 

 spectra were con5)uted. Intervals were selected for averaging to include 

 all intervals containing a reasonable number of cases and all high wave 

 intervals regardless of the number of cases. The minimum number of cases 

 considered for low and moderate waves was 10 at most locations. The wave 

 height above which all intervals were considered varies between locations. 

 The average spectrum for a height-period interval was obtained by aver- 

 aging energy values in each band. The standard deviation of spectral 

 energy values about the mean for each band was also computed. Averages 

 and standard deviations of spectra for the locations listed in Table 1 

 are discussed in the next section. 



It is important to note that the spectrum for any individual record 

 never duplicates the mean spectrum. Any application which is sensitive 

 to the precise form of the spectrum should make use of individual spectra 

 rather than the mean spectrum. One reasonable approach might be to select 

 a small random sample of spectra from each group, apply each spectrum to 



25 



