Plate 
technique employed, and that a similarity spectrum which is the same 
in field and laboratory is a physical possibility (see also Plate and 
Nath (1969) ). For the discussion in this and the following section, an 
equilibrium spectrum of universal shape and in the form given by 
Mitsuyasu is assumed to exist. 
The spectral shapes shown in Fig. 3 are averaged in the sense 
that the 'humps'"' in the experimental spectra have been removed by 
drawing a smooth curve through them. There is no requirement that 
a similarity spectrum must be a monotonous function, and the multiple 
peaks observed in the experimental spectra will, since they occur at 
multiples of w_, occur again in the similarity spectrum. I have shown 
in a recent paper (Plate (1971) ) that the concept of such a similarity 
spectrum leads naturally to an explanation of the overshoot phenomenon, 
But the significance of the similarity spectrum goes much further. For 
if the spectra are similar at all frequencies, then the measurement of 
the properties at one particular frequency suffices to fully specify the 
spectrum, This is an important conclusion, for it permits to reconcile 
the spectrum approach with the older dominant wave approach. Ifa 
single wave component suffices to describe the spectrum, why not use 
the dominant wave for this component ? 
This proposition is almost obvious, and yet, there is a very 
fundamental objection to it. For it is in general not permissible to 
identify component waves and physical waves. A physical wave isa 
water surface contour, while the component wave is a Fourier compo- 
nent. Consequently the former arises from a superposition of many of 
the latter. Fortunately, the sharp peak in the spectrum indicates that 
component waves of appreciable magnitudes are clustered around On 3 
and as a consequence are only weakly affected by components at higher 
or lower frequencies. This is of course reflected in the observed wave 
pattern. Since the dominant waves are waves that belong into a narrow 
band of components with frequency near w,, it follows that the domi- 
nant wave, the component wave at wy, and the higher maxima of the 
time series correspond to very nearly the same thing. In particular, 
one can assume that the highest n out of m waves below some number 
n<Nno all are dominant waves with frequency mn: 
Civil Engineers and Oceanographers are accustomed to use 
the average height of the highest one-third of observed waves as a 
measure of the wave height of the dominant wave. This usage reflects 
the observation that these waves remain well defined and do not vary 
greatly in length or period or progression speed. One therefore can 
1376 
