299 



had wave lengths of 1 - 3 cm. Their direction of propagation was primarily- 

 normal to the wind direction. As the wind speed increased, the ripples 

 initially present became larger in amplitude and height. -Under the action 

 of the steady air motion, the waves traveled downstream at an increasing 

 speed, growing in amplitude and length. For wind speeds in the range 

 Uoc = 3-6 mps, significant waves were observed to run with crests approximately 

 normal to the wind direction, with smooth windward surfaces, and rippled lee- 

 ward surfaces. Above 6 mps, capillary ripples were noted on both the wind- 

 ward and the leeward sides of the significant waves. At any given point 

 downstream from the inlet, groups of 5-20 small gravity waves of nearly the 

 same period passed by. These groups were separated by relatively calm regions 

 of small ripples having varied periods. The existence of groups of waves 

 separated by relatively calm water is probably rela-ted to interference between 

 different components of the wave train, giving an appearance of "beats . " 



The growth with fetch of waves in the channel is reflected in twQ. charac- 

 teristic lengths, the standard deviation a and the wave length X 

 The increase with F and Uoo of a and X is shown in Figure 8. The effect 

 of depth is also shown in the drawing. Decrease in depth tends to reduce the 

 wave length, and the standard deviation of the (larger) waves generated at 

 higher wind speeds . Our data for o and A were compared to those reported 

 by Sibul (1955) • For a given value of Uqc ^nd d, the results of both these 

 studies appeared to be essentially the same . 



Two characteristic velocities are associated with the water motion_^ These 

 are the surface velocity Uq, and the phase speed of sigaif icant waves Cg . The 

 change with F, Uqc and d of these properties, is shown in Figure 9- For a 

 given wind speed, the surface drift remains nearly constant ovg.r the range of 

 d shown, except near the ends of the channel. The wave speed Ce is approxi- 

 mately independent of depth down to ^.1 cm, but it increases with both Ucx) 

 and F. 



Keulegan (1951) found that the ratio of the drift velocity Uq to the 

 wind speed could be correlated with the Reynolds number Re^ = n^d/ v t^, 

 where v is the viscosity of the wa-ter. In his calculations, Keulegan 

 used an air speed averaged over the cross section of his channel, Uavg* Good-win 



(1965) found that Uavg " 0-^5 Uqo for the data in the CSU channel. Using this 

 relation, the values of Uq have been plotted with Rea as shown in Figure 10. 

 The drift velocities found in this study are correlated satisfactorily in terms 

 of Red' Our data fall about 30 percent lower than Keulegan 's curve for wavy 

 water. The difference between these -two studies may be accounted for in three 

 ways. First, if it is assumed that the mass flow of water and air are related 

 to each other and not the velocities, the momentum ratio, PwUo/Pa^avg 

 should be used in this correlation. Keulegan 's data were taken at sea level 

 while the CSU measurements were made at nearly I8OO m altitude. If our data 

 are correc-ted for the decrease in air density with altitude, they will fall 

 about 12 percent below Keulegan ' s curve . The remainder of the difference 

 between these experiments may be related to (a) the effect of non-uniformities 

 in Uq in the y-direction resulting from the nature of the air flow shown in 

 Figure 5, and (b) the fact that Keulegan used a value of Uq averaged over the 

 length of his channel while our values of Uq are taken locally. The effect 

 of (b) should be small, however, since uq varies little with fetch. 



