somewhere near the trough, and the wind direction may he reversed. 

 Since wind profile measurements near the surface have been taken 

 only at wind speeds of, approximately, less than 10 m/sec, the 

 rather short period wave motion passing by underneath the cup ane- 

 mometers "speeds up" the anemometers every time a crest passes. 

 The inertia of the anemometers does not allow slowing down over 

 the troughs, and therefore, almost maximum wind speeds are observed 

 instead of average values. When used for stress computations, too 

 small values of the stress are obviously derived by this procedure. 

 The "critical wind speed" at 7 m/sec as discussed by W. Munk (194-7) 

 can be explained as a result of the inaccuracy of the data obtained 

 by wind profile measurements (Neumann, 195D. 



Another more direct method of deriving the wind stress at the 

 sea surface is based on observations of the steady state sea sur- 

 face slope in which the water is piled up due to wind ("wind set 

 up"). This method is the oldest, and it was first used by W. Ekman 

 in 1905. Since that time much more data on the 'wind set up" in 

 different seas and lakes have accumulated, and with these data, the 

 method was reexamined and critically applied by the author in 1948. 



The stress, T , may be expressed in terms of a resistance co- 

 efficient, y 2 > which is defined by 



T= p« Y 2 W 2 , (31) 



where p 1 is the density of the air, and W is the mean wind speed at 

 a given distance above the sea surface. In practice, this distance 



is given by the "anemometer height", about 10 m. The question is 



2 

 whether y is a constant, or whether it depends on the wind speed 



itself. In the case of a flow over a solid surface with well de- 

 fined and constant roughness conditions, such as grass land or snow, 



38 



