b. The range of possible wave period variation within a single low- 

 frequency band translates into significantly different harbor responses in 

 the numerical model and, presumably, the field. 



The most straightforward change is to use a finer spectral bandwidth, though 

 finer bands have the undesirable consequence of lower confidence levels. SIO 

 provided CERC with one month (January 1994) of detailed line spectral coeffi- 

 cients to explore alternatives (Thompson 1995). The effect of bandwidth on the 

 T p estimate is illustrated in Figure 13 using data from the array (line spectra from 

 all four sensors averaged together). This record was selected because there was 

 an exceptional level of long wave energy in the harbor. Bandwidth is expressed 

 in the figure in terms of the number of spectral lines. The standard SIO proce- 

 dure for the Kahului Harbor gages gives 82 lines per band. Peak period esti- 

 mates are quite variable. For this particular case, T p estimated by the standard 

 procedure is 14.29 sec while the "true" peak period (middle of the scatter) 

 appears to be around 15 sec. The sawtooth shape of the plotted data arises 

 because the main energy concentration slowly marches toward shorter period as 

 bandwidth increases. Eventually, the band preceding the main energy extends 

 far enough to encompass that energy, and peak period abruptly shifts to the 

 center of that band. 



The artificial variability induced by bandwidth can be reduced by using 

 overlapping bands to identify a T^ The approach is to select a bandwidth and 

 identify T p . Then the bands (keeping the same bandwidth) are shifted a fixed 

 number of lines toward higher frequency (shorter period) and T p is again 

 estimated. The bands are shifted repeatedly and the final estimate of 7^ is based 

 on whichever band gives the very highest energy. The whole process can be 

 repeated with different choices of bandwidth to examine this effect as well. 

 Results with a two-line shift show a significantly reduced scatter relative to the 

 nonoverlapping approach (Figure 14). Thus the overlapping bands allow a more 

 refined estimate of T p . 



Two other cases in January 1994 corresponding to high levels of long wave 

 energy in the harbor were examined using the same overlapping band approach. 

 Peak period for 20 Jan 94 (1314) appears to be well-estimated by both the over- 

 lapping approach and the standard approach (Table 4). However, this case has a 

 relatively short T p and broad energy spectrum. The T p for 31 Jan 94 (0719) is 

 around 18 sec by the overlapping approach and 20 sec by the standard approach. 

 The standard analysis is not sufficiently discriminating for this case. 



Chapter 2 Field Wave Measurements 1 7 



