jgg MALKUS [chap. 4 



estimates differing by only 2° in direction and about 25% in magnitude. Even 

 the crude estimates of stress from the resultant wind and the mean wind speed 

 are little different from the rest. This is to be expected in tropical regions where 

 surface-wind steadiness (resultant divided by average speed times one hundred) 

 is high, so that the winds are nearly always from the prevailing direction. If we 

 assume that the direct summation method using the latest (solid line, Fig. 6) 

 determinations of C£) is the most correct stress, we may analyze the departures 

 therefrom obtained by the several wind-rose methods, recalling that the sample 

 is too small really to expect normal distributions. The good agreement of the 

 Scripps Atlantic technique may be regarded as accidental ; it is, in fact, their 

 crudest approach since it lumps together Beaufort numbers 1-3 into one 

 average wind speed. The result entitled "modified Scripps Atlantic" avoided 

 this, but used the more recent cd (solid) curve. The Scripps Pacific method 

 takes a Gaussian speed distribution, but a cd from the dashed line in Fig. 6. 

 In this set of observations, a 25% reduction of stress magnitude (but no change 

 in direction) results from each modification. 



The next important question concerns the resemblance of the resultant 

 stress for the period of this expedition to the local April mean from the Scripps 

 charts. The latter shows only a circle north of Puerto Rico, or stress less than 

 0.6 dynes/cm2. The Scripps method was, however, apphed to the local April 

 wind rose {U.S. Hydrographic Office Pilot Charts) shown in Fig. 44b, which 

 gave a shearing stress of about 0.3 dynes/cm2 from 065°. Therefore, although 

 the expedition studied a period of stronger-than-average trades (see Bunker 

 et al., 1949; also Malkus, 1958), its resultant shearing stress is in the same 

 direction and only twice as big as the April average. Similar variations may be 

 expected throughout the steady and relatively undisturbed trade regime. 



Quite a different situation is encountered over the high-latitude oceans, where 

 the significant stress is produced mainly in transient cyclonic storms. There 

 wind steadiness is low and the mean picture may say little about what might 

 be met in a given week or month. We have obtained wind data for the period 

 February 26-March 19, 1960, from Weather Ship C, stationed at 52° 45'N, 

 35° 30'W, both from the crew's log, which estimates velocity in Beaufort number 

 from the sea state, and from the special weather observers' data in knots from 

 anemometers. Even the latter are less reliable than winds from a research 

 vessel, since the weather ship was usually steaming, as well as pitching and 

 rolling in heavy seas. Resultant stresses computed from (17) by the different 

 methods are summarized in Table XV. A histogram of air-sea temperature 

 difference distribution is plotted in Fig. 44f to give an idea of the atmospheric 

 stability. There were only 18% of the hourly observations which showed 

 7^Q_ J'^> 3.0°C, indicating instability (roughly), and only 20% where the 

 ocean was colder than the air. Under these conditions stress estimations from 

 (17) should be correct to within 50% and all sensible methods in Table XV agree 

 with each other to better than this leeway. Even the use of the crew's log wind 

 data gives a stress vector differing only 15° in direction and 15% in magnitude 

 from the direct summation methods, suggesting the feasibility of obtaining 



