storm during a 2-hr collection interval). If these assumptions are violated, 

 results are suspect. Hence, there are three measurement problems for which to 

 check: malfunctioning gages, temporal non-stationarity , and spatial in- 

 homogeneity . 



98. A spatially inhomogeneous wave field is one where properties change 

 substantially over some region of interest, in the present case over the 

 length of the linear array. If wave direction differs from one end of the 

 array to the other, the cross -spectra of signals between pairs of equally 

 separated gages will be different. This condition arises because the time 

 difference of wave forms passing one relative to the second of a pair of gages 

 is directly related to wave direction. A preliminary examination of cross- 

 spectral densities at redundant gage spacings under a variety of wave fields 

 measured with the FRF linear array indicated no significant errors due to 

 spatial inhomogeneity . Hence, it has been assumed in all subsequent analysis 

 that wave conditions are homogeneous . 



99 . Of the remaining problems of malfunctioning gages and temporal non- 

 stationarity , the former has been found to be the most severe. An error- 

 checking computer program has been written to identify failed gages so they 

 could be eliminated from further analysis. The error-checking principle 

 employed is quite simple. Since multiple gages were deployed in an (assumed) 

 homogeneous sea, certain statistical properties of data from each gage should 

 be identical. Intercomparison of data from all gages can then reveal inopera- 

 tive gages. Two types of properties were used. One was integral properties , 

 requiring the summing of data values; the other was point properties , which 

 examined extreme points in a set of time series. 



100. Integral properties were mean value, standard deviation, skewness, 

 kurtosis, and trend. The first four of these are conventional time-series 

 parameters, defined in any statistics text. Trend is the difference of the 

 end values of a straight line fitted through the data. It gives a measure of 

 the bulk (low- frequency) rate of change of a measured property. 



101. The procedure was to compute each of the integral properties for 

 each of the four, sequential 4,096-point records for each gage in a collec- 

 tion. All 10 gages were used since the tenth gage was sensing virtually the 

 same environment. Division into four records provided a sequence in time of 

 these statistical properties. For each of the four records and for each 

 property there were then 10 estimates of that property, one for each gage. 



39 



