theory. For each pair of gauges, raw cross-spectral estimates are then com- 

 puted, ensemble averaged, and band averaged to create a smoothed cross- 

 spectrum for each gauge pair. For each smoothed frequency, a matrix is 

 formed of the cross-spectra for all possible pairs of gauges. This matrix is 

 then fed to the directional estimation algorithm. 



Wave energy directional distributions are computed using the Iterative 

 Maximum Likelihood Estimator (IMLE), which was derived by Pawka (1982, 

 1983), and found to be a substantial improvement on the Maximum Likeli- 

 hood Estimator (MLE) derived by Davis and Regier (1977). Specifically, the 

 algorithm is Equations 1 - 4 in the paper by Pawka (1983) with convergence 

 exponent /3 = 1 , convergence coefficient 7 = 5, and a maximum of 30 itera- 

 tions. 



In practice, all gauges of the FRF array are not used for all wave frequen- 

 cies. Gauges with large separations are used for long, low-frequency waves. 

 Closely spaced gauges are used for shorter waves to avoid loss of coherence 

 for these waves at large spatial separations. Details of array sub-configura- 

 tions and their associated frequencies are given by Long and Oltman-Shay 

 (1991). Tests of the FRF array with artificial data indicate a directional reso- 

 lution of 5 deg. This means it can resolve the peak of a single mode direc- 

 tional distribution to within 5 deg, or distinguish the peaks of a bimodal distri- 

 bution when separated by 15 deg or more. 



Anemometer 



The reference anemometer used in this study was mounted atop the FRF 

 laboratory building (Figure 2) at an elevation of 19.5 m above mean sea level. 

 It has a nominal speed accuracy of ±0.5 m/sec and direction accuracy of 

 ±3 deg. Speed and direction data are collected on separate channels, and 

 results from each channel are simply averaged (not vector averaged). The 

 proximity of the anemometer to the FRF building is known to cause low speed 

 readings on some occasions in very high winds (Long and Hubertz 1988), but 

 it is considered sufficiently reliable for the generally moderate winds encoun- 

 tered in this study. Though other anemometers have been deployed for short- 

 er periods at the FRF, this one was consistently in use for the duration of this 

 study. It follows that this gauge provides the most consistent local represen- 

 tation of the larger scale wind field. 



Data Sampling Schemes 



With few exceptions, data used in this report were obtained following the 

 standard FRF sampling scheme. The pattern is to collect four contiguous 

 records of 34-min duration (4,096 data points at a 2-Hz sampling rate) starting 

 at 0100, 0700, 1300, and 1900 Eastern Standard Time (EST). When the 

 characteristic wave height fl^ exceeds 2 m at a reference wave gauge (usual- 

 ly gauge #1 of the directional array), additional collections are made starting 

 at 0400, 1000, 1600, and 2200 EST. This form of sampling gives good data 



Chapter 2 Measurement Scheme 



