coverage of high energy conditions, but has a tendency to miss some of the 

 details of initial transient conditions at the onset of a storm. 



The anemometer data presented in this report are based on a subset of the 

 overall FRF database. Wind speeds and directions are deduced from single 

 34-min records with start times of 0100, 0700, 1300, and 1900 EST. These 

 particular records are the most reliable, having been checked by hand for use 

 in the FRF annual reports (Leffler et al. 1992, 1993). The result is a very 

 regular sampling of local winds, with gaps only for system maintenance or 

 repair. 



All possible data were processed from the directional wave gauge, so that 

 normally there are samples from all the routine start times listed above. Four 

 records of 34 min each (2 hr 16 min or 16,384 data points at 2 Hz) for each 

 gauge were used in cross-spectral computation to ensure high confidence in 

 the resulting cross-spectral matrix that is transformed in the IMLE algorithm. 

 Fifteen half-overlapping segments of 2,048 points were processed for ensem- 

 ble averages, and 10 subsequent frequency bands were averaged for a final 

 resolution bandwidth A/ = 0.00977 Hz and a minimum of 160 degrees of 

 freedom (excluding any gain in degrees of freedom from the lapped segments) 

 in the final cross-spectral estimates. While this method gives very good re- 

 sults in estimates of S{f,6), it has the effect of averaging all wave processes 

 for the 2 hr 16 min records so that highly transient processes are smoothed 

 somewhat. Results are not invalid, just rather strongly filtered. Shorter 

 records can be used in future work if brief transients are important. 



The bulk of S(f,6) results were obtained by the method just described, but 

 there were, on occasion, minor variations in sampling times and rates. Dur- 

 ing the SUPERDUCK experiment in September and October 1986 (Crowson 

 et al. 1988), data collection start times were adjusted to conform to high and 

 low tides, and so did not conform to the FRF normal collection pattern. 

 Furthermore, a total of seven 34-min records were obtained in each collection. 

 To maximize the number of S(f,6) observations during SUPERDUCK, re- 

 cords 1 - 4 from a collection were processed, and then records 4-7 were 

 processed, so that record 4 was used twice, but observations were still almost 

 independent. Time series of parameters from these spectra thus appear rather 

 oddly spaced, with pairs of samples appearing with 1 hr 42 min separations. 

 During the DELILAH experiment of October 1990 (Birkemeier et al. 1991), 

 data were collected for an additional 34 min and the sampling rate for ail data 

 was 4 Hz instead of 2 Hz. Thus, a collection had 10 records of 4,096 data 

 points (about 17 min) each. For the data used here, wind speeds and direc- 

 tions during DELILAH were averaged for only 17 min instead of the usual 

 34 min. Directional array data were still processed in 2 hr 16 min total re- 

 cord lengths, but, in smoothing, 20 bands were averaged instead of 10, result- 

 ing in the same final resolution bandwidth (but more degrees of freedom). 



Chapter 2 Measurement Scheme 



