Detection and measurement of wave breaking - Breaking and Dissipation of Ocean Surface Waves - page 87

This diagnosis of breaking waves was found to be approximately 70-75% accurate once
the optimum threshold had been determined.
The classification-accuracy analysis detailed in Manasseh et al. ( 2006 ) should be applied
in further field experiments, covering a wider range of wind-wave environments than were
available in Lake George. If in each future field experiment, synchronised surface video
and underwater audio records were made with the same equipment and settings (for exam-
ple, with the hydrophone the same distance below the surface and the camera the same
distance above), the classification-accuracy analysis could be repeated on several data sets
from a wider range of sea-state environments than studied here. If the classification accu-
racy of 75% is not worsened, it would suggest the discriminant recommended (in sound
pressure amplitude at the hydrophone) has a universality.
The method was then used for detailed analysis of wave-breaking properties across the
spectrum. To obtain wave-breaking probabilities of individual waves at different frequen-
cies, a zero-crossing analysis was applied. From the time series of surface elevations, the
period of each wave was calculated as follows. Times when the surface elevation crossed
the mean or 'zero' level were noted. Two consecutive zero up-crossings were analysed
and the time of the troughs preceding and following them were recorded. The difference
between the trough times was taken as the period of that wave T , giving its frequency.
Figure 3.9 (top panel) shows a 30 s section of the surface elevation data of record 5
( Table 5.2 ) used to calculate wave frequencies.
In Figure 3.9 , limitations of the zero-crossing analysis at small scales are quite obvious.
These limitations and other issues with the zero-crossing analysis are discussed in the
Appendix of Manasseh et al. ( 2006 ). A superior alternative for higher-frequency breaking,
which also preserves the wave shape contrary to the standard Fourier-based bandpassing
procedures, would be a riding-wave removal (RWR) method ( Schulz , 2009 ).
The synchronous passive acoustic wave-breaking data was then combined with the
surface-elevation data and the zero-crossing analysis. Such bubble-detection events are
shown, for the same time series, in the bottom panel of Figure 3.9 . Occasional events
which would correspond to negative surface elevations were excluded from the analysis to
avoid possible ambiguity in detecting wave breakers when those events happened close to
wave troughs. For each acoustically determined breaker, the frequency of the wave at the
same time was extracted. The total number of breaking waves n

( f ) was found for each of

the calculated frequencies f .
When applied to real field data, a breaking-probability distribution could thus be obtained;
this will be described in Section 5.3 . This is the rate of occurrence of wave-breaking events
at different wave scales. Therefore, the method provides spectral distribution of the break-
ing probability (2.4) rather than just the frequency of occurrence of dominant breakers
(2.3) like the spectrograms. For the dominant breaking, the two techniques were com-
pared and showed effectively the same breaking count, provided that the assumed dominant
frequency band is

35 f p (2.8) . This comparison is shown in Figure 3.10 .
With support from a separate laboratory experiment, the estimated bubble size was
argued to be dependent on the severity of wave breaking and thus to provide information

f p =± 0 .