Jin Wu 
1968). The average values of the wavelength, calculated from the 
measured period by using the dispersion relation for small-amplitu- 
de deep-water waves, are shown in figure 6. A general observation 
is that small-amplitude capillary waves are generated at very low 
wind velocities (U, < 2 m/sec). Rhombic wave cells are observed 
before the wind boundary layer becomes fully rough (2 m/sec <U5—3 
m/sec); the waves forming cells are short gravity waves. As the 
wind velocity further increases, the propagation of waves becomes 
more and more along the direction of the wind. Parasitic capillaries 
are formed in front of the gravity wave crests immediately following 
the transition of the wind boundary layer from smooth to rough (U, = 
3 m/sec), the water surface is generally smooth elsewhere. As the 
wind velocity increases passing 9 m/sec, ripples are seen covering 
the water surface and breaking is observed along the wave crests, 
which span transversely across the tank. Various stages of wave 
growth with the wind velocity are indicated in figure 6. 
IV. SURFACE SLOPES 
IV. 1 Distribution of surface slopes 
For each wind velocity, the number of light pulses in 10-se- 
cond intervals is counted electrically with the optical instrument set 
at various angles of inclination. Each light pulse represents the occur- 
rence of a particular water-surface slope, whose inclination from the 
horizontal is the same as that of the instrument from the vertical. 
More than 30 sets of readings are recorded for each inclination, of 
which the average value as well as the standard deviation from the 
average are plotted versus the angle of inclination in figure 7. As 
shown in the lower right corner of figure 7, slopes with their normals 
pointing up-wind are considered to be negative, whereas those pointing 
down-wind are positive. A short vertical line accompanying each data 
point indicates the value of the standard deviation of the data from the 
average. The reliability of the data is indicated by the small values 
of the standard deviations. 
The data points shown in figure 7 are seen to follow essential- 
ly a gaussian distribution. The area defined by the data points is first 
integrated by means of Simpson's rule to determine the median value. 
Around this median value, the data points for each wind velocity are then 
normalized. The normal distribution curve, found on the basis of least- 
square curve fitting, is drawn as the continuous line shown in figure 
7. It is now seen that the nearly normally distributed data are skewed 
(maximum and median values at angles other than 0°) at medium wind 
velocity and peaked (maximum values beyond the gaussian curve) at 
1442 
