69 

 Sheppard and Webb, 1956; Sheppard, 1958) • It is found to increase with 

 windspeed and with decreasing Richardson niiinber - that is, the drag of the 

 sea on the air appears to be greater at lower stability with the same 

 windspeed. The drag coefficient roughly doubles as the wind increases from 

 2 to ik meters per sec. Normal tropical variations in Richardson number 

 contribute only about one-fourth this much variation. 



Keeping these problems in mind, we turn to our first topic, namely 

 the role of tropical sea-air fluxes in the planetary circulations. To 

 examine this, we need the climatological picture of these fluxes - the 

 magnitude of evaporation and sensible heat exchange - on annual global 

 maps, and analysis of regional and seasonal variations. 



The classical maps of Jacobs (1951) are sho\m again in Figures 1 and 

 2, Malkus (1962) and Garstang (196'+) have compared these distributions with 

 the later results of Budyko (1956) who also used the transfer formulas but 

 does not divulge his data sources or method of analysis. In looking at 

 these charts, we must keep in mind one further important limitation and 

 that is the data problem. Even supposing that the transfer equations were 

 exactly correct with an exactly known coefficient, for good results we 

 would need a measurement network of air temperature and humidity, sea 

 temperature and windspeed reporting every few hours. Then we should make 

 the multiplications required by equation (l) from each set of data and 

 average these products over month, season or year. 



Of course, this is a visionary goal. Climatological mean values have 

 to be plugged directly into the formulas and clearly this can lead to errors 

 if there are correlations between air-sea property difference and windspeed - 

 if, for example, the air temperature commonly drops in storm situations. 

 This points the finger directly at synoptic disturbances. 



Malkus (1962) and Garstang (196^4-) have made several case studies where 

 the transfer calculation from fairly long-period means could be compared 

 with averages of frequent measurements of the input into equation (l) from 

 research vessels or Weather Ships. Suffice it to say here that in the strong 

 and steady trades the correlation is unimportant, but wherever disturbances 

 are predominant, such as in the equatorial trough, the error can become 

 quite large. Garstang 's (196^) results indicate that there it may be 

 considerably larger than the earlier estimates by Malkus (1962). 



Figures 1 and 2 show by and large the expected flux distributions, 

 with some curious features, such as higher transfers in the Pacific than 

 in the Atlantic, which exhibits negative sensible heat flux off North 

 Africa. Budyko 's (1956) values are greater in the Atlantic, suggesting that 

 another scale of fluctuations has perhaps distorted the picture c Jacob's 

 calculations were made using values at 1200 GOT, which is near midday in 

 the Atlantic area and at night in the Pacific . We do not know what observa- 

 tions Budyko used, but the diurnal transfer cycle that we shall describe 

 later suggests that he may have used these at 0000 GOT. In all trajisfer 





