215 



q'w', u'w', and s'w' depends upon the situation. The interval must be long 

 enough that consistent values of the mean products are obtained, but yet not so 

 long that time trends of the fluxes of the different quantities are eliminated. The 

 quantity T itself is one of the unknowns to be ascertained experimentally.* 



One of the prime factors to be considered in the design of an instrument 

 which will yield satisfactory information for the determination of eddy flux in ad- 

 dition to stability is its time response characteristics. The instrument must be 

 capable of responding to fluctuations having periods much less than the interval 

 T. More specifically, the instrument should have a T factor (i.e., the time in- 

 terval corresponding to a response of the instrument of 1/e of a sudden change 

 impressed upon it) which is considerably less than the period of those fluctua- 

 tions associated with the largest amplitude. Teniperature is one quantity for 

 which suitable instruments can be devised. However, velocity and especially 

 humidity or salinity present a more formidable problem to the oceanographic or 

 meteorological instrument designer. In some cases, the problem of measuring 

 such quantities can be reduced to one of measuring a temperature, or an elec- 

 tromotive force. Table I contains a brief summary of present instruments or 

 measuring elements capable of yielding a continuous record of the quantities'*^ 

 q, u or w, and s. Devices for the measurement of radiant flux of energy are 

 also mentioned in this table. 



It has been pointed out by Swinbank (1951-b) that, in the interpretation of 

 observed values from any continuous record, the amplitude distortion as well as 

 the phase shift introduced by the imperfect response of the measuring and re- 

 cording system must be taken into account. For example, if the response char- 

 acteristics of the humidity device are vastly different from that of the velocity 

 measuring equipment, then the mean products of these quantities will be in er- 

 ror due to the differential distortion and nonsimultaneity introduced by phase 

 shift. This effect will be most pronounced for the short period fluctuations. 



The above source of error can be minimized by reducing the response fac- 

 tor T to as low a value as is possible. An alternative procedure is to adjust the 

 measuring systems for a given pair of parameters to be utilized in a flux deter- 

 mination such that each system has the same T* factor, provided that this factor 

 is less than the period of the greatest amplitudes of fluctuations. 



Table I is by no means intended as a comprehensive review of possible in- 

 struments for measurement of the fluctuations desired, but at least it indicates 

 the possibility of eddy flux measurement over the water. Indeed, Swinbank has 

 demonstrated that such measurements are possible over land. 



* - It is surmised but not definitely established that a "plateau" in the realm 

 of microscopic turbulence exists which would allow the establishment of aver- 

 ages which are not too sensitive to the interval of averaging (within certain 

 bounds). Indeed it is just such a "plateau", between the microscopic and micro- 

 scopic, which underlies the science of thermodynamics. 



