SECT. 2] LAKGE-SOALE INTERACTIONS 111 



speed and property difference should be obtained, the product made of their 

 product with the cd read from Fig. 6 and the average of this sum taken 

 as the average transfer at a given locahty. Numerous local measurement 

 points should have observations available to be dealt with in this manner in 

 order to draw isopleths of exchange to compare with radiation balances, etc. 

 Clearly this situation is not and never will be realized, and we must usually 

 calculate fluxes from equations (17) to (19) using climatological (monthly or 

 longer) averages of property differences and wind speed, and a mean value of 

 Cj) appropriate to the latter. We may analyze the errors in these determinations 

 by breaking down the factors entering an equation of the type (18) into long- 

 period averages, designated by a bar, and hourly departures therefrom, 

 designated by primes, namely, 



^ 2 pCyjCp + c' d){AT + AT'){Ua^ 



N 



[18b) 



pCpCD A T Ua + pCp[CD{A T'u'a) + Ua(c'j) A T'] 



+ AT{c'jju'a) + c'i,AT'u'al (18c) 



where To—Ta has been written AT, variations in p are assumed negligible, 

 and N is the number of hourly observations, which are indicated in (18a) by 

 unprimed unbarred quantities. 



The first term on the right in (18c) is the heat flux calculated from the long- 

 period mean values of the observables and we wish to examine the closeness of 

 its approximation to Qs : namely, to find out how good is the relation 



Q's = il pCpCD ATua)IN ;^ pCpCo ATua. (18d) 



This depends upon the size of all the neglected terms in the brackets in (18c). If 

 the relationship between c/> and Ua is satisfactorily given by Fig. 6, then the 

 adequacy of using climatological average data in determinations of air-sea ex- 

 change depends entirely upon the correlation between air-sea property dif- 

 ference and wind speed, and the standard deviations of each. 



A sample examination of these correlations and comparisons of the type 

 indicated in (18d) has been made using the ship observations of the Wyman- 

 Woodcook expedition to the Caribbean in April, 1946 (Wyman et al., 1946). The 

 results are presented in Table II. In considering sensible heat flux Qs, a small 

 but statistically significant negative correlation ( — 0.22) was found between air- 

 sea temperature difference and wind speed, suggesting that the ocean surface 

 layers are allowed to warm up in the absence of stirring by strong winds. The 

 larger negative correlation between the wind speed and surface temperature 

 confirms this. However, the standard deviations in AT and Ua were such that 

 Qs computed in the two ways of (18d) differed by less than 7%. A similar 

 formulation and analysis for latent heat flux Qe was performed with nearly 



