second moment about the mean (variance) and then edits the data by checking 
for "jumps," "spikes," and points exceeding the voltage limit of the gage. A 
jump is defined as a data value greater than five standard deviations from the 
previous data value, whereas a spike is a data value more than five standard 
deviations from the mean. If less than five consecutive jumps or spikes are 
found, the program linearly interpolates between acceptable data and replaces 
the erroneous data values. The editing stops if the program finds more than 
five consecutive jumps or spikes, or more than a total of 100 bad points, or 
the variance of the voltage is below 1 x 10° squared volts. The statistics 
and diagnostics from the analysis are saved. 
35. Sea surface energy spectra are computed from the edited time 
series. Spectral estimates are computed from smaller data segments obtained 
by dividing the 4,096—point record into several 512—point segments. The 
estimates are then ensemble—averaged to produce a more accurate spectrum. 
These data segments are overlapped by 50 percent (known as the Welch (1967) 
method) which has been shown to produce better statistical properties than 
nonoverlapped segments. The mean and linear trends are removed from each 
segment prior to spectral analysis. To reduce side—lobe leakage in the spec— 
tral estimates, a data window was applied. The first and last 10 percent of 
data points were multiplied by a cosine bell (Bingham, Godfrey, and Tukey 
1967). Spectra were computed from each segment with a discrete Fast Fourier 
Transform and then ensemble—-averaged. Sea surface spectra from subsurface 
pressure gages were obtained by applying the linear wave theory transfer 
function. 
36. Unless otherwise stated, wave height in this report refers to the 
energy—based parameter H,, defined as four times the zeroth moment wave 
height of the estimated sea surface spectrum (i.e., four times the square root 
of the variance) computed from the spectrum passband. Energy computations 
from the spectra are limited to a passband between 0.05 and 0.50 Hz for sur- 
face gages and between 0.05 Hz and a high-frequency cutoff for subsurface 
gages. This high-frequency limit is imposed to eliminate aliased energy and 
noise measurements from biasing the computation of H,, and is defined as the 
frequency where the linear theory transfer function is less than 0.1 (spectral 
values are multiplied by 100 or more). Smoother and more statistically 
significant spectral estimates are obtained by band—averaging contiguous 
spectral components (three components are averaged per band, producing a 
25 
