significant wave height was over 2 feet and the discrepancy of individual 

 standard deviations from their mean was 3 percent or less . Plots of the 

 high-resolution spectra for these observations are in Appendix E. The 

 vertical lines represent the energy contribution at each spectral period. 

 The background level for each observation was estimated from the minima 

 between 25 and 7 seconds in the spectra. Spectral periods in this range 

 with energy content . above twice the estimated background energy were 

 identified. Contributions to the energy satisfying this criterion at 

 adjacent spectral periods were considered as arising from the same wave 

 train. The number of adjacent periods in each train was used as a measure 

 of the speotval width of the train. The energy had to be above the chosen 

 level at all five gages for the spectral period to be included in the 

 group. The spectral period among these showing maximum energy was taken 

 as the "period" of the wave train. 



Directions were computed at all the spectral periods in each train 

 for the 10 arrays. The total spread among these directions was found, and 

 an average total spread was computed for the trains having the same spec- 

 tral width. The same was done for the computed directional spread at the 

 period of the train. 



Twenty-five percent of the identified wave trains had total direc- 

 tional (computed) spreads above 100° and were not considered further. 

 For 89 percent of the discarded trains, the period of the train was under 

 9.4 seconds. Thus, all trains with periods under 9.4 seconds were dis- 

 carded. 



Results for the different spectral widths for the trains retained 

 (280) are shown in Table 6. The second column in the table gives the 

 average total directional spread for the corresponding spectral width; 

 the third column gives the average directional spread at the period of 

 the wave train. The last column gives the number of wave trains having 

 the spectral width in the first column. 



These results indicate that the total directional spread increased 

 with frequency width. Narrow peaks consisting of from one to three 

 spectral periods are most frequent, and the spread in the direction at 

 the period of the train remains relatively constant. Since the average 

 spread for narrow-banded trains (width <0.003 hertz) is 21.8°, it is 

 expected that three-gage arrays cannot yield directional results to any 

 better accuracy. The mathematical exercise in Appendix C shows that 

 array dimension is a limiting factor as to the shortest period for which 

 some directional discrimination may be expected. An important factor in 

 the validity of the directional result is the spectral structure of the 

 wave train involved. Only in very special circumstances will the quanti- 

 ties involved in equation C-10 (App. C) combine to give better results. 



There are various possible explanations for the large spreads observed 

 in the directional results from field records. For the long-crested wave 

 model to be strictly applicable, it is important that: 



29 



