voltages were stored for each gage as four records of 4,096 points (34 min 8 

 sec). This record division was used later in data quality control. 



94. A time line of satisfactory data collection start times is shown in 

 Figure 4. Most of the first year's data collection coincided with routine 

 daily FRF observations (Birkemeier et al . 1985) at 0100, 0700, 1300, and 1900 

 Eastern Standard Time (EST) under low-energy conditions. Additional daily 

 collections at 0400, 1000, 1600, and 2200 EST occurred under high-energy 

 conditions, defined as times when H^^ at a reference gage exceeded a pre- 

 defined level or when specifically prescribed by an investigator. 



95. At the beginning of the collection year, during September and 

 October 1986, the SUPERDUCK experiment (Crowson et al . 1988) was conducted. 

 During this period, collections were timed with tidal extremes, and data were 

 gathered for seven records of 4,096 points for each gage. These collections 

 were processed by considering records 1 to 4 as one sample and records 4 to 7 

 as a second sample, even though the two samples overlapped by one record. 

 This gave a closely spaced sampling of almost independent observations during 

 SUPERDUCK. 



96. Gaps in the data stream in Figure 4 are the result of several 

 causes. Short gaps were caused mostly by computer downtime or gage wiring 

 problems such as ground loops or loose connections. Longer gaps during 

 SUPERDUCK were caused by problems associated with installing a new gage. 

 Serious long-term data collection for this report did not begin until late in 

 January 1987, so there is a 3 -month gap following SUPERDUCK. For the remain- 

 der of the year, data coverage was good. A total of 1,046 frequency-direction 

 spectra were measured. 



Error Checking 



97. Conversion of observed data to frequency- direction spectral 

 estimates requires computing products of frequency spectra from the nine gages 

 in the array. This means that data quality must be exceptionally high so that 

 spiky or drifty data from one gage do not contaminate results in spectral 

 multiplication with results from the other eight gages. Additionally, the 

 linear array analysis programs are based on the hypotheses that the wave field 

 is homogeneous along the array (i.e., the same wave trains occur at all gages) 

 and stationary in time (e.g., does not go from a flat sea to a full-blown 



37 



