Because of the uncertainty concerning the validity of the assumptions 

 made earlier with regard to the characteristics of the sample distributions, 

 the nonparametric Kolmogorov-Smirnov (K-S) two-sample test (Siegel, 1956) 

 was used to retest the null hypothesis that these data samples are measure- 

 ments of the same environment. This test, which would be a much stronger 

 one if the assumptions of independence and normality were not entirely valid, 

 is based on the assumption that the cumulative frequency distributions of both 

 samples will agree closely if both are from the same population, because the 

 samples should show only random deviations from the population distribution. 

 Wide differences in the sample cumulative frequency distributions, on the 

 other hand, would suggest that the samples came from different populations 

 and the null hypothesis would be rejected. 



For the K-S test, each sample is grouped into the same set of intervals 

 (S(x)) and the maximum difference D between like intervals is the value tested: 



D = Maximum 



S(x)i - S(x)2 



The critical value of D, D^,, is the value that must be equaled or exceeded 

 in order for the null hypothesis to be rejected. According to Smirnov 

 (Siegel, 1956), D^ is calculated from 



n^ + ^2 



Dp = 1.36 for a = 0.05. 



c n^ • n2 



Because this test requires that arbitrary intervals be established, and 

 because too few intervals will tend to mask the maximum difference, the 

 minimum number of intervals that would show the maximum differences in like 

 intervals was determined. The best interval length was determined by study 

 of intervals of 0.1°C, 0.2°C, 0.25°C, and 0.5°C based on sample boom and 

 shipboard data, and the 0.2°C interval length was chosen as the most effi- 

 cient for use in this study. 



5. RESULTS 



5.1 useless Oaeanographer 



On the Ooeanographev the shipboard observations were made aft of the 

 exhaust stacks near the rawinsonde inflation shelter 220 ft from the bow 

 where the boom was located. Figure 8 shows the average differences between 

 the shipboard and boom temperatures at scheduled rawinsonde release times. 

 When the wind was not blowing across or along the ship toward the boom, the 

 shipboard temperatures were warmer than the boom temperatures between 1300 

 and 0300 GMT, and the boom temperatures generally were warmer but occasionally 

 very slightly cooler than the temperatures observed on board the ship between 

 0430 and 1200 GMT. Therefore, these data were divided into 0400-1259 and 

 1300-0359 GMT subgroups. After similar analysis of cases when the wind was 

 blowing across or along the ship toward the boom, the data were divided into 

 1000-2159 and 2200-0959 GMT subgroups. 



